Patent Grant
11269808
Embodiments of the disclosure relate to stateless data ingestion over a network protocol. More specifically, one embodiment of the disclosure relates to a computerized method for storing ingested data in a persistent storage prior to minimally processing and storing the ingested data as event-based data.
Information technology (IT) environments can include diverse types of data systems that store large amounts of diverse data types generated by numerous devices. For example, a big data ecosystem may include databases such as MySQL and Oracle databases, cloud computing services such as Amazon web services (AWS), and other data systems that store passively or actively generated data, including machine-generated data (“machine data”). The machine data can include performance data, diagnostic data, or any other data that can be analyzed to diagnose equipment performance problems, monitor user interactions, and to derive other insights.
The large amount and diversity of data systems containing large amounts of structured, semi-structured, and unstructured data relevant to any search query can be massive, and continues to grow rapidly. This technological evolution can give rise to various challenges in relation to managing, understanding and effectively utilizing the data. To reduce the potentially vast amount of data that may be generated, some data systems pre-process data based on anticipated data analysis needs. In particular, specified data items may be extracted from the generated data and stored in a data system to facilitate efficient retrieval and analysis of those data items at a later time. At least some of the remainder of the generated data is typically discarded during pre-processing.
However, storing massive quantities of minimally processed or unprocessed data (collectively and individually referred to as “raw data”) for later retrieval and analysis is becoming increasingly more feasible as storage capacity becomes more inexpensive and plentiful. In general, storing raw data and performing analysis on that data later can provide greater flexibility because it enables an analyst to analyze all of the generated data instead of only a fraction of it.
Minimally processing the raw data may include segmenting the raw data into predetermined sized blocks and annotating each block with metadata. In particular, one metadata field may be a source type. When these blocks are searched, the source type metadata field may be the basis for selecting one or more configuration files that determine extraction rules. Thus, in order to provide accurate search results, it is imperative that the source type is assigned, and done so accurately, so as to determine the one or more applicable configuration files, and ultimately, the appropriate extraction rules.
A common trend in the development of computer-implemented technology is standardization, whereby different systems are enabled to interoperate by a standard protocol, interface, or the like. For example, the majority of the current world-wide-web operates according to a version of the hypertext transport protocol (HTTP), a protocol known in the art. Given the prevalence of HTTP, it is often desirable to transmit data from data sources to an analysis system via HTTP.
The present disclosure is illustrated by way of example, and not limitation, in the figures of the accompanying drawings, in which like reference numerals indicate similar elements and in which:
Embodiments are described herein according to the following outline:
1.0. General Overview
2.0. Operating Environment
1.0. General Overview
Modern data centers and other computing environments can comprise anywhere from a few host computer systems to thousands of systems configured to process data, service requests from remote clients, and perform numerous other computational tasks. During operation, various components within these computing environments often generate significant volumes of machine data. Machine data is any data produced by a machine or component in an information technology (IT) environment and that reflects activity in the IT environment. For example, machine data can be raw machine data that is generated by various components in IT environments, such as servers, sensors, routers, mobile devices, Internet of Things (IoT) devices, etc. Machine data can include system logs, network packet data, sensor data, application program data, error logs, stack traces, system performance data, etc. In general, machine data can also include performance data, diagnostic information, and many other types of data that can be analyzed to diagnose performance problems, monitor user interactions, and to derive other insights.
A number of tools are available to analyze machine data. In order to reduce the size of the potentially vast amount of machine data that may be generated, many of these tools typically pre-process the data based on anticipated data-analysis needs. For example, pre-specified data items may be extracted from the machine data and stored in a database to facilitate efficient retrieval and analysis of those data items at search time. However, the rest of the machine data typically is not saved and is discarded during pre-processing. As storage capacity becomes progressively cheaper and more plentiful, there are fewer incentives to discard these portions of machine data and many reasons to retain more of the data.
This plentiful storage capacity is presently making it feasible to store massive quantities of minimally processed machine data for later retrieval and analysis. In general, storing minimally processed machine data and performing analysis operations at search time can provide greater flexibility because it enables an analyst to search all of the machine data, instead of searching only a pre-specified set of data items. This may enable an analyst to investigate different aspects of the machine data that previously were unavailable for analysis.
However, analyzing and searching massive quantities of machine data presents a number of challenges. For example, a data center, servers, or network appliances may generate many different types and formats of machine data (e.g., system logs, network packet data (e.g., wire data, etc.), sensor data, application program data, error logs, stack traces, system performance data, operating system data, virtualization data, etc.) from thousands of different components, which can collectively be very time-consuming to analyze. In another example, mobile devices may generate large amounts of information relating to data accesses, application performance, operating system performance, network performance, etc. There can be millions of mobile devices that report these types of information.
These challenges can be addressed by using an event-based data intake and query system, such as the SPLUNK® ENTERPRISE system developed by Splunk Inc. of San Francisco, Calif. The SPLUNK® ENTERPRISE system is the leading platform for providing real-time operational intelligence that enables organizations to collect, index, and search machine data from various websites, applications, servers, networks, and mobile devices that power their businesses. The data intake and query system is particularly useful for analyzing data which is commonly found in system log files, network data, and other data input sources. Although many of the techniques described herein are explained with reference to a data intake and query system similar to the SPLUNK® ENTERPRISE system, these techniques are also applicable to other types of data systems.
In the data intake and query system, machine data are collected and stored as “events”. An event comprises a portion of machine data and is associated with a specific point in time. The portion of machine data may reflect activity in an IT environment and may be produced by a component of that IT environment, where the events may be searched to provide insight into the IT environment, thereby improving the performance of components in the IT environment. Events may be derived from “time series data,” where the time series data comprises a sequence of data points (e.g., performance measurements from a computer system, etc.) that are associated with successive points in time. In general, each event has a portion of machine data that is associated with a timestamp that is derived from the portion of machine data in the event. A timestamp of an event may be determined through interpolation between temporally proximate events having known timestamps or may be determined based on other configurable rules for associating timestamps with events.
In some instances, machine data can have a predefined format, where data items with specific data formats are stored at predefined locations in the data. For example, the machine data may include data associated with fields in a database table. In other instances, machine data may not have a predefined format (e.g., may not be at fixed, predefined locations), but may have repeatable (e.g., non-random) patterns. This means that some machine data can comprise various data items of different data types that may be stored at different locations within the data. For example, when the data source is an operating system log, an event can include one or more lines from the operating system log containing machine data that includes different types of performance and diagnostic information associated with a specific point in time (e.g., a timestamp).
Examples of components which may generate machine data from which events can be derived include, but are not limited to, web servers, application servers, databases, firewalls, routers, operating systems, and software applications that execute on computer systems, mobile devices, sensors, Internet of Things (IoT) devices, etc. The machine data generated by such data sources can include, for example and without limitation, server log files, activity log files, configuration files, messages, network packet data, performance measurements, sensor measurements, etc.
The data intake and query system uses a flexible schema to specify how to extract information from events. A flexible schema may be developed and redefined as needed. Note that a flexible schema may be applied to events “on the fly,” when it is needed (e.g., at search time, index time, ingestion time, etc.). When the schema is not applied to events until search time, the schema may be referred to as a “late-binding schema.”
During operation, the data intake and query system receives machine data from any type and number of sources (e.g., one or more system logs, streams of network packet data, sensor data, application program data, error logs, stack traces, system performance data, etc.). The system parses the machine data to produce events each having a portion of machine data associated with a timestamp. The system stores the events in a data store. The system enables users to run queries against the stored events to, for example, retrieve events that meet criteria specified in a query, such as criteria indicating certain keywords or having specific values in defined fields. As used herein, the term “field” refers to a location in the machine data of an event containing one or more values for a specific data item. A field may be referenced by a field name associated with the field. As will be described in more detail herein, a field is defined by an extraction rule (e.g., a regular expression) that derives one or more values or a sub-portion of text from the portion of machine data in each event to produce a value for the field for that event. The set of values produced are semantically-related (such as IP address), even though the machine data in each event may be in different formats (e.g., semantically-related values may be in different positions in the events derived from different sources).
As described above, the system stores the events in a data store. The events stored in the data store are field-searchable, where field-searchable herein refers to the ability to search the machine data (e.g., the raw machine data) of an event based on a field specified in search criteria. For example, a search having criteria that specifies a field name “UserID” may cause the system to field-search the machine data of events to identify events that have the field name “UserID.” In another example, a search having criteria that specifies a field name “UserID” with a corresponding field value “12345” may cause the system to field-search the machine data of events to identify events having that field-value pair (e.g., field name “UserID” with a corresponding field value of “12345”). Events are field-searchable using one or more configuration files associated with the events. Each configuration file includes one or more field names, where each field name is associated with a corresponding extraction rule and a set of events to which that extraction rule applies. The set of events to which an extraction rule applies may be identified by metadata associated with the set of events. For example, an extraction rule may apply to a set of events that are each associated with a particular host, source, or source type. When events are to be searched based on a particular field name specified in a search, the system uses one or more configuration files to determine whether there is an extraction rule for that particular field name that applies to each event that falls within the criteria of the search. If so, the event is considered as part of the search results (and additional processing may be performed on that event based on criteria specified in the search). If not, the next event is similarly analyzed, and so on.
As noted above, the data intake and query system utilizes a late-binding schema while performing queries on events. One aspect of a late-binding schema is applying extraction rules to events to extract values for specific fields during search time. More specifically, the extraction rule for a field can include one or more instructions that specify how to extract a value for the field from an event. An extraction rule can generally include any type of instruction for extracting values from events. In some cases, an extraction rule comprises a regular expression, where a sequence of characters form a search pattern. An extraction rule comprising a regular expression is referred to herein as a regex rule. The system applies a regex rule to an event to extract values for a field associated with the regex rule, where the values are extracted by searching the event for the sequence of characters defined in the regex rule.
In the data intake and query system, a field extractor may be configured to automatically generate extraction rules for certain fields in the events when the events are being created, indexed, or stored, or possibly at a later time. Alternatively, a user may manually define extraction rules for fields using a variety of techniques. In contrast to a conventional schema for a database system, a late-binding schema is not defined at data ingestion time. Instead, the late-binding schema can be developed on an ongoing basis until the time a query is actually executed. This means that extraction rules for the fields specified in a query may be provided in the query itself, or may be located during execution of the query. Hence, as a user learns more about the data in the events, the user can continue to refine the late-binding schema by adding new fields, deleting fields, or modifying the field extraction rules for use the next time the schema is used by the system. Because the data intake and query system maintains the underlying machine data and uses a late-binding schema for searching the machine data, it enables a user to continue investigating and learn valuable insights about the machine data.
In some embodiments, a common field name may be used to reference two or more fields containing equivalent and/or similar data items, even though the fields may be associated with different types of events that possibly have different data formats and different extraction rules. By enabling a common field name to be used to identify equivalent and/or similar fields from different types of events generated by disparate data sources, the system facilitates use of a “common information model” (CIM) across the disparate data sources (further discussed with respect to
2.0. Operating Environment
The networked computer system 100 comprises one or more computing devices. These one or more computing devices comprise any combination of hardware and software configured to implement the various logical components described herein. For example, the one or more computing devices may include one or more memories that store instructions for implementing the various components described herein, one or more hardware processors configured to execute the instructions stored in the one or more memories, and various data repositories in the one or more memories for storing data structures utilized and manipulated by the various components.
In some embodiments, one or more client devices 102 are coupled to one or more host devices 106 and a data intake and query system 108 via one or more networks 104. Networks 104 broadly represent one or more LANs, WANs, cellular networks (e.g., LTE, HSPA, 3G, and other cellular technologies), and/or networks using any of wired, wireless, terrestrial microwave, or satellite links, and may include the public Internet.
2.1. Host Devices
In the illustrated embodiment, a system 100 includes one or more host devices 106. Host devices 106 may broadly include any number of computers, virtual machine instances, and/or data centers that are configured to host or execute one or more instances of host applications 114. In general, a host device 106 may be involved, directly or indirectly, in processing requests received from client devices 102. Each host device 106 may comprise, for example, one or more of a network device, a web server, an application server, a database server, etc. A collection of host devices 106 may be configured to implement a network-based service. For example, a provider of a network-based service may configure one or more host devices 106 and host applications 114 (e.g., one or more web servers, application servers, database servers, etc.) to collectively implement the network-based application.
In general, client devices 102 communicate with one or more host applications 114 to exchange information. The communication between a client device 102 and a host application 114 may, for example, be based on the Hypertext Transfer Protocol (HTTP) or any other network protocol. Content delivered from the host application 114 to a client device 102 may include, for example, HTML documents, media content, etc. The communication between a client device 102 and host application 114 may include sending various requests and receiving data packets. For example, in general, a client device 102 or application running on a client device may initiate communication with a host application 114 by making a request for a specific resource (e.g., based on an HTTP request), and the application server may respond with the requested content stored in one or more response packets.
In the illustrated embodiment, one or more of host applications 114 may generate various types of performance data during operation, including event logs, network data, sensor data, and other types of machine data. For example, a host application 114 comprising a web server may generate one or more web server logs in which details of interactions between the web server and any number of client devices 102 is recorded. As another example, a host device 106 comprising a router may generate one or more router logs that record information related to network traffic managed by the router. As yet another example, a host application 114 comprising a database server may generate one or more logs that record information related to requests sent from other host applications 114 (e.g., web servers or application servers) for data managed by the database server.
2.2. Client Devices
Client devices 102 of
2.3. Client Device Applications
In some embodiments, each client device 102 may host or execute one or more client applications 110 that are capable of interacting with one or more host devices 106 via one or more networks 104. For instance, a client application 110 may be or comprise a web browser that a user may use to navigate to one or more websites or other resources provided by one or more host devices 106. As another example, a client application 110 may comprise a mobile application or “app.” For example, an operator of a network-based service hosted by one or more host devices 106 may make available one or more mobile apps that enable users of client devices 102 to access various resources of the network-based service. As yet another example, client applications 110 may include background processes that perform various operations without direct interaction from a user. A client application 110 may include a “plug-in” or “extension” to another application, such as a web browser plug-in or extension.
In some embodiments, a client application 110 may include a monitoring component 112. At a high level, the monitoring component 112 comprises a software component or other logic that facilitates generating performance data related to a client device's operating state, including monitoring network traffic sent and received from the client device and collecting other device and/or application-specific information. Monitoring component 112 may be an integrated component of a client application 110, a plug-in, an extension, or any other type of add-on component. Monitoring component 112 may also be a stand-alone process.
In some embodiments, a monitoring component 112 may be created when a client application 110 is developed, for example, by an application developer using a software development kit (SDK). The SDK may include custom monitoring code that can be incorporated into the code implementing a client application 110. When the code is converted to an executable application, the custom code implementing the monitoring functionality can become part of the application itself.
In some embodiments, an SDK or other code for implementing the monitoring functionality may be offered by a provider of a data intake and query system, such as a system 108. In such cases, the provider of the system 108 can implement the custom code so that performance data generated by the monitoring functionality is sent to the system 108 to facilitate analysis of the performance data by a developer of the client application or other users.
In some embodiments, the custom monitoring code may be incorporated into the code of a client application 110 in a number of different ways, such as the insertion of one or more lines in the client application code that call or otherwise invoke the monitoring component 112. As such, a developer of a client application 110 can add one or more lines of code into the client application 110 to trigger the monitoring component 112 at desired points during execution of the application. Code that triggers the monitoring component may be referred to as a monitor trigger. For instance, a monitor trigger may be included at or near the beginning of the executable code of the client application 110 such that the monitoring component 112 is initiated or triggered as the application is launched, or included at other points in the code that correspond to various actions of the client application, such as sending a network request or displaying a particular interface.
In some embodiments, the monitoring component 112 may monitor one or more aspects of network traffic sent and/or received by a client application 110. For example, the monitoring component 112 may be configured to monitor data packets transmitted to and/or from one or more host applications 114. Incoming and/or outgoing data packets can be read or examined to identify network data contained within the packets, for example, and other aspects of data packets can be analyzed to determine a number of network performance statistics. Monitoring network traffic may enable information to be gathered particular to the network performance associated with a client application 110 or set of applications.
In some embodiments, network performance data refers to any type of data that indicates information about the network and/or network performance. Network performance data may include, for instance, a Uniform Resource Locator (URL) requested, a connection type (e.g., HTTP, HTTPS, etc.), a connection start time, a connection end time, an HTTP status code, request length, response length, request headers, response headers, connection status (e.g., completion, response time(s), failure, etc.), and the like. Upon obtaining network performance data indicating performance of the network, the network performance data can be transmitted to a data intake and query system 108 for analysis.
Upon developing a client application 110 that incorporates a monitoring component 112, the client application 110 can be distributed to client devices 102. Applications generally can be distributed to client devices 102 in any manner, or they can be pre-loaded. In some cases, the application may be distributed to a client device 102 via an application marketplace or other application distribution system. For instance, an application marketplace or other application distribution system might distribute the application to a client device based on a request from the client device to download the application.
Examples of functionality that enables monitoring performance of a client device are described in U.S. patent application Ser. No. 14/524,748, entitled “UTILIZING PACKET HEADERS TO MONITOR NETWORK TRAFFIC IN ASSOCIATION WITH A CLIENT DEVICE”, filed on 27 Oct. 2014, and which is hereby incorporated by reference in its entirety for all purposes.
In some embodiments, the monitoring component 112 may also monitor and collect performance data related to one or more aspects of the operational state of a client application 110 and/or client device 102. For example, a monitoring component 112 may be configured to collect device performance information by monitoring one or more client device operations, or by making calls to an operating system and/or one or more other applications executing on a client device 102 for performance information. Device performance information may include, for instance, a current wireless signal strength of the device, a current connection type and network carrier, current memory performance information, a geographic location of the device, a device orientation, and any other information related to the operational state of the client device.
In some embodiments, the monitoring component 112 may also monitor and collect other device profile information including, for example, a type of client device, a manufacturer and model of the device, versions of various software applications installed on the device, and so forth.
In general, a monitoring component 112 may be configured to generate performance data in response to a monitor trigger in the code of a client application 110 or other triggering application event, as described above, and to store the performance data in one or more data records. Each data record, for example, may include a collection of field-value pairs, each field-value pair storing a particular item of performance data in association with a field for the item. For example, a data record generated by a monitoring component 112 may include a “networkLatency” field (not shown in the Figure) in which a value is stored. This field indicates a network latency measurement associated with one or more network requests. The data record may include a “state” field to store a value indicating a state of a network connection, and so forth for any number of aspects of collected performance data.
2.4. Data Server System
Each data source 202 broadly represents a distinct source of data that can be consumed by system 108. Examples of a data sources 202 include, without limitation, data files, directories of files, data sent over a network, event logs, registries, etc.
During operation, the forwarders 204 identify which indexers 206 receive data collected from a data source 202 and forward the data to the appropriate indexers. Forwarders 204 can also perform operations on the data before forwarding, including removing extraneous data, detecting timestamps in the data, parsing data, indexing data, routing data based on criteria relating to the data being routed, and/or performing other data transformations.
In some embodiments, a forwarder 204 may comprise a service accessible to client devices 102 and host devices 106 via a network 104. For example, one type of forwarder 204 may be capable of consuming vast amounts of real-time data from a potentially large number of client devices 102 and/or host devices 106. The forwarder 204 may, for example, comprise a computing device which implements multiple data pipelines or “queues” to handle forwarding of network data to indexers 206. A forwarder 204 may also perform many of the functions that are performed by an indexer. For example, a forwarder 204 may perform keyword extractions on raw data or parse raw data to create events. A forwarder 204 may generate time stamps for events. Additionally or alternatively, a forwarder 204 may perform routing of events to indexers 206. Data store 208 may contain events derived from machine data from a variety of sources all pertaining to the same component in an IT environment, and this data may be produced by the machine in question or by other components in the IT environment.
2.5. Cloud-Based System Overview
The example data intake and query system 108 described in reference to
In one embodiment, to provide an alternative to an entirely on-premises environment for system 108, one or more of the components of a data intake and query system instead may be provided as a cloud-based service. In this context, a cloud-based service refers to a service hosted by one more computing resources that are accessible to end users over a network, for example, by using a web browser or other application on a client device to interface with the remote computing resources. For example, a service provider may provide a cloud-based data intake and query system by managing computing resources configured to implement various aspects of the system (e.g., forwarders, indexers, search heads, etc.) and by providing access to the system to end users via a network. Typically, a user may pay a subscription or other fee to use such a service. Each subscribing user of the cloud-based service may be provided with an account that enables the user to configure a customized cloud-based system based on the user's preferences.
In some embodiments, a cloud-based data intake and query system 306 may comprise a plurality of system instances 308. In general, each system instance 308 may include one or more computing resources managed by a provider of the cloud-based system 306 made available to a particular subscriber. The computing resources comprising a system instance 308 may, for example, include one or more servers or other devices configured to implement one or more forwarders, indexers, search heads, and other components of a data intake and query system, similar to system 108. As indicated above, a subscriber may use a web browser or other application of a client device 302 to access a web portal or other interface that enables the subscriber to configure an instance 308.
Providing a data intake and query system as described in reference to system 108 as a cloud-based service presents a number of challenges. Each of the components of a system 108 (e.g., forwarders, indexers, and search heads) may at times refer to various configuration files stored locally at each component. These configuration files typically may involve some level of user configuration to accommodate particular types of data a user desires to analyze and to account for other user preferences. However, in a cloud-based service context, users typically may not have direct access to the underlying computing resources implementing the various system components (e.g., the computing resources comprising each system instance 308) and may desire to make such configurations indirectly, for example, using one or more web-based interfaces. Thus, the techniques and systems described herein for providing user interfaces that enable a user to configure source type definitions are applicable to both on-premises and cloud-based service contexts, or some combination thereof (e.g., a hybrid system where both an on-premises environment, such as SPLUNK® ENTERPRISE, and a cloud-based environment, such as SPLUNK CLOUD□, are centrally visible).
2.6. Searching Externally-Archived Data
The search head 210 of the data intake and query system receives search requests from one or more client devices 404 over network connections 420. As discussed above, the data intake and query system 108 may reside in an enterprise location, in the cloud, etc.
The search head 210 analyzes the received search request to identify request parameters. If a search request received from one of the client devices 404 references an index maintained by the data intake and query system, then the search head 210 connects to one or more indexers 206 of the data intake and query system for the index referenced in the request parameters. That is, if the request parameters of the search request reference an index, then the search head accesses the data in the index via the indexer. The data intake and query system 108 may include one or more indexers 206, depending on system access resources and requirements. As described further below, the indexers 206 retrieve data from their respective local data stores 208 as specified in the search request. The indexers and their respective data stores can comprise one or more storage devices and typically reside on the same system, though they may be connected via a local network connection.
If the request parameters of the received search request reference an external data collection, which is not accessible to the indexers 206 or under the management of the data intake and query system, then the search head 210 can access the external data collection through an External Result Provider (ERP) process 410. An external data collection may be referred to as a “virtual index” (plural, “virtual indices”). An ERP process provides an interface through which the search head 210 may access virtual indices.
Thus, a search reference to an index of the system relates to a locally stored and managed data collection. In contrast, a search reference to a virtual index relates to an externally stored and managed data collection, which the search head may access through one or more ERP processes 410, 412.
The search head 210 may spawn a single ERP process in response to multiple virtual indices referenced in a search request, or the search head may spawn different ERP processes for different virtual indices. Generally, virtual indices that share common data configurations or protocols may share ERP processes. For example, all search query references to a Hadoop file system may be processed by the same ERP process, if the ERP process is suitably configured. Likewise, all search query references to a SQL database may be processed by the same ERP process. In addition, the search head may provide a common ERP process for common external data source types (e.g., a common vendor may utilize a common ERP process, even if the vendor includes different data storage system types, such as Hadoop and SQL). Common indexing schemes also may be handled by common ERP processes, such as flat text files or Weblog files.
The search head 210 determines the number of ERP processes to be initiated via the use of configuration parameters that are included in a search request message. Generally, there is a one-to-many relationship between an external results provider “family” and ERP processes. There is also a one-to-many relationship between an ERP process and corresponding virtual indices that are referred to in a search request. For example, using RDBMS, assume two independent instances of such a system by one vendor, such as one RDBMS for production and another RDBMS used for development. In such a situation, it is likely preferable (but optional) to use two ERP processes to maintain the independent operation as between production and development data. Both of the ERPs, however, will belong to the same family, because the two RDBMS system types are from the same vendor.
The ERP processes 410, 412 receive a search request from the search head 210. The search head may optimize the received search request for execution at the respective external virtual index. Alternatively, the ERP process may receive a search request as a result of analysis performed by the search head or by a different system process. The ERP processes 410, 412 can communicate with the search head 210 via conventional input/output routines (e.g., standard in/standard out, etc.). In this way, the ERP process receives the search request from a client device such that the search request may be efficiently executed at the corresponding external virtual index.
The ERP processes 410, 412 may be implemented as a process of the data intake and query system. Each ERP process may be provided by the data intake and query system, or may be provided by process or application providers who are independent of the data intake and query system. Each respective ERP process may include an interface application installed at a computer of the external result provider that ensures proper communication between the search support system and the external result provider. The ERP processes 410, 412 generate appropriate search requests in the protocol and syntax of the respective virtual indices 414, 416, each of which corresponds to the search request received by the search head 210. Upon receiving search results from their corresponding virtual indices, the respective ERP process passes the result to the search head 210, which may return or display the results or a processed set of results based on the returned results to the respective client device.
Client devices 404 may communicate with the data intake and query system 108 through a network interface 420, e.g., one or more LANs, WANs, cellular networks, intranetworks, and/or internetworks using any of wired, wireless, terrestrial microwave, satellite links, etc., and may include the public Internet.
The analytics platform utilizing the External Result Provider process described in more detail in U.S. Pat. No. 8,738,629, entitled “External Result Provided Process For RetriEving Data Stored Using A Different Configuration Or Protocol”, issued on 27 May 2014, U.S. Pat. No. 8,738,587, entitled “PROCESSING A SYSTEM SEARCH REQUEST BY RETRIEVING RESULTS FROM BOTH A NATIVE INDEX AND A VIRTUAL INDEX”, issued on 25 Jul. 2013, U.S. patent application Ser. No. 14/266,832, entitled “PROCESSING A SYSTEM SEARCH REQUEST ACROSS DISPARATE DATA COLLECTION SYSTEMS”, filed on 1 May 2014, and U.S. Pat. No. 9,514,189, entitled “PROCESSING A SYSTEM SEARCH REQUEST INCLUDING EXTERNAL DATA SOURCES”, issued on 6 Dec. 2016, each of which is hereby incorporated by reference in its entirety for all purposes.
2.6.1. ERP Process Features
The ERP processes described above may include two operation modes: a streaming mode and a reporting mode. The ERP processes can operate in streaming mode only, in reporting mode only, or in both modes simultaneously. Operating in both modes simultaneously is referred to as mixed mode operation. In a mixed mode operation, the ERP at some point can stop providing the search head with streaming results and only provide reporting results thereafter, or the search head at some point may start ignoring streaming results it has been using and only use reporting results thereafter.
The streaming mode returns search results in real time, with minimal processing, in response to the search request. The reporting mode provides results of a search request with processing of the search results prior to providing them to the requesting search head, which in turn provides results to the requesting client device. ERP operation with such multiple modes provides greater performance flexibility with regard to report time, search latency, and resource utilization.
In a mixed mode operation, both streaming mode and reporting mode are operating simultaneously. The streaming mode results (e.g., the machine data obtained from the external data source) are provided to the search head, which can then process the results data (e.g., break the machine data into events, timestamp it, filter it, etc.) and integrate the results data with the results data from other external data sources, and/or from data stores of the search head. The search head performs such processing and can immediately start returning interim (streaming mode) results to the user at the requesting client device; simultaneously, the search head is waiting for the ERP process to process the data it is retrieving from the external data source as a result of the concurrently executing reporting mode.
In some instances, the ERP process initially operates in a mixed mode, such that the streaming mode operates to enable the ERP quickly to return interim results (e.g., some of the machined data or unprocessed data necessary to respond to a search request) to the search head, enabling the search head to process the interim results and begin providing to the client or search requester interim results that are responsive to the query. Meanwhile, in this mixed mode, the ERP also operates concurrently in reporting mode, processing portions of machine data in a manner responsive to the search query. Upon determining that it has results from the reporting mode available to return to the search head, the ERP may halt processing in the mixed mode at that time (or some later time) by stopping the return of data in streaming mode to the search head and switching to reporting mode only. The ERP at this point starts sending interim results in reporting mode to the search head, which in turn may then present this processed data responsive to the search request to the client or search requester. Typically, the search head switches from using results from the ERP's streaming mode of operation to results from the ERP's reporting mode of operation when the higher bandwidth results from the reporting mode outstrip the amount of data processed by the search head in the streaming mode of ERP operation.
A reporting mode may have a higher bandwidth because the ERP does not have to spend time transferring data to the search head for processing all the machine data. In addition, the ERP may optionally direct another processor to do the processing.
The streaming mode of operation does not need to be stopped to gain the higher bandwidth benefits of a reporting mode; the search head could simply stop using the streaming mode results—and start using the reporting mode results—when the bandwidth of the reporting mode has caught up with or exceeded the amount of bandwidth provided by the streaming mode. Thus, a variety of triggers and ways to accomplish a search head's switch from using streaming mode results to using reporting mode results may be appreciated by one skilled in the art.
The reporting mode can involve the ERP process (or an external system) performing event breaking, time stamping, filtering of events to match the search query request, and calculating statistics on the results. The user can request particular types of data, such as if the search query itself involves types of events, or the search request may ask for statistics on data, such as on events that meet the search request. In either case, the search head understands the query language used in the received query request, which may be a proprietary language. One exemplary query language is Splunk Processing Language (SPL) developed by the assignee of the application, Splunk Inc. The search head typically understands how to use that language to obtain data from the indexers, which store data in a format used by the SPLUNK® Enterprise system.
The ERP processes support the search head, as the search head is not ordinarily configured to understand the format in which data is stored in external data sources such as Hadoop or SQL data systems. Rather, the ERP process performs that translation from the query submitted in the search support system's native format (e.g., SPL if SPLUNK® ENTERPRISE is used as the search support system) to a search query request format that will be accepted by the corresponding external data system. The external data system typically stores data in a different format from that of the search support system's native index format, and it utilizes a different query language (e.g., SQL or MapReduce, rather than SPL or the like).
As noted, the ERP process can operate in the streaming mode alone. After the ERP process has performed the translation of the query request and received raw results from the streaming mode, the search head can integrate the returned data with any data obtained from local data sources (e.g., native to the search support system), other external data sources, and other ERP processes (if such operations were required to satisfy the terms of the search query). An advantage of mixed mode operation is that, in addition to streaming mode, the ERP process is also executing concurrently in reporting mode. Thus, the ERP process (rather than the search head) is processing query results (e.g., performing event breaking, timestamping, filtering, possibly calculating statistics if required to be responsive to the search query request, etc.). It should be apparent to those skilled in the art that additional time is needed for the ERP process to perform the processing in such a configuration. Therefore, the streaming mode will allow the search head to start returning interim results to the user at the client device before the ERP process can complete sufficient processing to start returning any search results. The switchover between streaming and reporting mode happens when the ERP process determines that the switchover is appropriate, such as when the ERP process determines it can begin returning meaningful results from its reporting mode.
The operation described above illustrates the source of operational latency: streaming mode has low latency (immediate results) and usually has relatively low bandwidth (fewer results can be returned per unit of time). In contrast, the concurrently running reporting mode has relatively high latency (it has to perform a lot more processing before returning any results) and usually has relatively high bandwidth (more results can be processed per unit of time). For example, when the ERP process does begin returning report results, it returns more processed results than in the streaming mode, because, e.g., statistics only need to be calculated to be responsive to the search request. That is, the ERP process doesn't have to take time to first return machine data to the search head. As noted, the ERP process could be configured to operate in streaming mode alone and return just the machine data for the search head to process in a way that is responsive to the search request. Alternatively, the ERP process can be configured to operate in the reporting mode only. Also, the ERP process can be configured to operate in streaming mode and reporting mode concurrently, as described, with the ERP process stopping the transmission of streaming results to the search head when the concurrently running reporting mode has caught up and started providing results. The reporting mode does not require the processing of all machine data that is responsive to the search query request before the ERP process starts returning results; rather, the reporting mode usually performs processing of chunks of events and returns the processing results to the search head for each chunk.
For example, an ERP process can be configured to merely return the contents of a search result file verbatim, with little or no processing of results. That way, the search head performs all processing (such as parsing byte streams into events, filtering, etc.). The ERP process can be configured to perform additional intelligence, such as analyzing the search request and handling all the computation that a native search indexer process would otherwise perform. In this way, the configured ERP process provides greater flexibility in features while operating according to desired preferences, such as response latency and resource requirements.
2.7. Data Ingestion
2.7.1. Input
At block 502, a forwarder receives data from an input source, such as a data source 202 shown in
At block 504, a forwarder or other system component annotates each block generated from the raw data with one or more metadata fields. These metadata fields may, for example, provide information related to the data block as a whole and may apply to each event that is subsequently derived from the data in the data block. For example, the metadata fields may include separate fields specifying each of a host, a source, and a source type related to the data block. A host field may contain a value identifying a host name or IP address of a device that generated the data. A source field may contain a value identifying a source of the data, such as a pathname of a file or a protocol and port related to received network data. A source type field may contain a value specifying a particular source type label for the data. Additional metadata fields may also be included during the input phase, such as a character encoding of the data, if known, and possibly other values that provide information relevant to later processing steps. In some embodiments, a forwarder forwards the annotated data blocks to another system component (typically an indexer) for further processing.
The data intake and query system allows forwarding of data from one data intake and query instance to another, or even to a third-party system. The data intake and query system can employ different types of forwarders in a configuration.
In some embodiments, a forwarder may contain the essential components needed to forward data. A forwarder can gather data from a variety of inputs and forward the data to an indexer for indexing and searching. A forwarder can also tag metadata (e.g., source, source type, host, etc.).
In some embodiments, a forwarder has the capabilities of the aforementioned forwarder as well as additional capabilities. The forwarder can parse data before forwarding the data (e.g., can associate a time stamp with a portion of data and create an event, etc.) and can route data based on criteria such as source or type of event. The forwarder can also index data locally while forwarding the data to another indexer.
2.7.2. Parsing
At block 506, an indexer receives data blocks from a forwarder and parses the data to organize the data into events. In some embodiments, to organize the data into events, an indexer may determine a source type associated with each data block (e.g., by extracting a source type label from the metadata fields associated with the data block, etc.) and refer to a source type configuration corresponding to the identified source type. The source type definition may include one or more properties that indicate to the indexer to automatically determine the boundaries within the received data that indicate the portions of machine data for events. In general, these properties may include regular expression-based rules or delimiter rules where, for example, event boundaries may be indicated by predefined characters or character strings. These predefined characters may include punctuation marks or other special characters including, for example, carriage returns, tabs, spaces, line breaks, etc. If a source type for the data is unknown to the indexer, an indexer may infer a source type for the data by examining the structure of the data. Then, the indexer can apply an inferred source type definition to the data to create the events.
At block 508, the indexer determines a timestamp for each event. Similar to the process for parsing machine data, an indexer may again refer to a source type definition associated with the data to locate one or more properties that indicate instructions for determining a timestamp for each event. The properties may, for example, instruct an indexer to extract a time value from a portion of data for the event, to interpolate time values based on timestamps associated with temporally proximate events, to create a timestamp based on a time the portion of machine data was received or generated, to use the timestamp of a previous event, or use any other rules for determining timestamps.
At block 510, the indexer associates with each event one or more metadata fields including a field containing the timestamp determined for the event. In some embodiments, a timestamp may be included in the metadata fields. These metadata fields may include any number of “default fields” that are associated with all events, and may also include one more custom fields as defined by a user. Similar to the metadata fields associated with the data blocks at block 504, the default metadata fields associated with each event may include a host, source, and source type field including or in addition to a field storing the timestamp.
At block 512, an indexer may optionally apply one or more transformations to data included in the events created at block 506. For example, such transformations can include removing a portion of an event (e.g., a portion used to define event boundaries, extraneous characters from the event, other extraneous text, etc.), masking a portion of an event (e.g., masking a credit card number), removing redundant portions of an event, etc. The transformations applied to events may, for example, be specified in one or more configuration files and referenced by one or more source type definitions.
As mentioned above, certain metadata, e.g., host 536, source 537, source type 538 and timestamps 535 can be generated for each event, and associated with a corresponding portion of machine data 539 when storing the event data in a data store, e.g., data store 208. Any of the metadata can be extracted from the corresponding machine data, or supplied or defined by an entity, such as a user or computer system. The metadata fields can become part of or stored with the event. Note that while the time-stamp metadata field can be extracted from the raw data of each event, the values for the other metadata fields may be determined by the indexer based on information it receives pertaining to the source of the data separate from the machine data.
While certain default or user-defined metadata fields can be extracted from the machine data for indexing purposes, all the machine data within an event can be maintained in its original condition. As such, in embodiments in which the portion of machine data included in an event is unprocessed or otherwise unaltered, it is referred to herein as a portion of raw machine data. In other embodiments, the port of machine data in an event can be processed or otherwise altered. As such, unless certain information needs to be removed for some reasons (e.g. extraneous information, confidential information), all the raw machine data contained in an event can be preserved and saved in its original form. Accordingly, the data store in which the event records are stored is sometimes referred to as a “raw record data store.” The raw record data store contains a record of the raw event data tagged with the various default fields.
In
In the example shown in
Event 534 is associated with an entry in a server error log, as indicated by “error.log” in the source column 537, that records errors that the server encountered when processing a client request. Similar to the events related to the server access log, all the raw machine data in the error log file pertaining to event 534 can be preserved and stored as part of the event 534.
Saving minimally processed or unprocessed machine data in a data store associated with metadata fields in the manner similar to that shown in
2.7.3. Indexing
At blocks 514 and 516, an indexer can optionally generate a keyword index to facilitate fast keyword searching for events. To build a keyword index, at block 514, the indexer identifies a set of keywords in each event. At block 516, the indexer includes the identified keywords in an index, which associates each stored keyword with reference pointers to events containing that keyword (or to locations within events where that keyword is located, other location identifiers, etc.). When an indexer subsequently receives a keyword-based query, the indexer can access the keyword index to quickly identify events containing the keyword.
In some embodiments, the keyword index may include entries for field name-value pairs found in events, where a field name-value pair can include a pair of keywords connected by a symbol, such as an equals sign or colon. This way, events containing these field name-value pairs can be quickly located. In some embodiments, fields can automatically be generated for some or all of the field names of the field name-value pairs at the time of indexing. For example, if the string “dest=10.0.1.2” is found in an event, a field named “dest” may be created for the event, and assigned a value of “10.0.1.2”.
At block 518, the indexer stores the events with an associated timestamp in a data store 208. Timestamps enable a user to search for events based on a time range. In some embodiments, the stored events are organized into “buckets,” where each bucket stores events associated with a specific time range based on the timestamps associated with each event. This improves time-based searching, as well as allows for events with recent timestamps, which may have a higher likelihood of being accessed, to be stored in a faster memory to facilitate faster retrieval. For example, buckets containing the most recent events can be stored in flash memory rather than on a hard disk. In some embodiments, each bucket may be associated with an identifier, a time range, and a size constraint.
Each indexer 206 may be responsible for storing and searching a subset of the events contained in a corresponding data store 208. By distributing events among the indexers and data stores, the indexers can analyze events for a query in parallel. For example, using map-reduce techniques, each indexer returns partial responses for a subset of events to a search head that combines the results to produce an answer for the query. By storing events in buckets for specific time ranges, an indexer may further optimize the data retrieval process by searching buckets corresponding to time ranges that are relevant to a query.
In some embodiments, each indexer has a home directory and a cold directory. The home directory of an indexer stores hot buckets and warm buckets, and the cold directory of an indexer stores cold buckets. A hot bucket is a bucket that is capable of receiving and storing events. A warm bucket is a bucket that can no longer receive events for storage but has not yet been moved to the cold directory. A cold bucket is a bucket that can no longer receive events and may be a bucket that was previously stored in the home directory. The home directory may be stored in faster memory, such as flash memory, as events may be actively written to the home directory, and the home directory may typically store events that are more frequently searched and thus are accessed more frequently. The cold directory may be stored in slower and/or larger memory, such as a hard disk, as events are no longer being written to the cold directory, and the cold directory may typically store events that are not as frequently searched and thus are accessed less frequently. In some embodiments, an indexer may also have a quarantine bucket that contains events having potentially inaccurate information, such as an incorrect time stamp associated with the event or a time stamp that appears to be an unreasonable time stamp for the corresponding event. The quarantine bucket may have events from any time range; as such, the quarantine bucket may always be searched at search time. Additionally, an indexer may store old, archived data in a frozen bucket that is not capable of being searched at search time. In some embodiments, a frozen bucket may be stored in slower and/or larger memory, such as a hard disk, and may be stored in offline and/or remote storage.
Moreover, events and buckets can also be replicated across different indexers and data stores to facilitate high availability and disaster recovery as described in U.S. Pat. No. 9,130,971, entitled “Site-Based Search Affinity”, issued on 8 Sep. 2015, and in U.S. patent Ser. No. 14/266,817, entitled “Multi-Site Clustering”, issued on 1 Sep. 2015, each of which is hereby incorporated by reference in its entirety for all purposes.
The data store 501 can correspond to a data store 208 that stores events managed by an indexer 206 or can correspond to a different data store associated with an indexer 206. In the illustrated embodiment, the data store 501 includes a _main directory 503 associated with a _main index and a _test directory 505 associated with a _test index. However, the data store 501 can include fewer or more directories. In some embodiments, multiple indexes can share a single directory or all indexes can share a common directory. Additionally, although illustrated as a single data store 501, it will be understood that the data store 501 can be implemented as multiple data stores storing different portions of the information shown in
In the illustrated embodiment of
In some embodiments, the inverted index 507A . . . 507B, and 509A . . . 509B can correspond to a distinct time-series bucket that is managed by the indexer 206 and that contains events corresponding to the relevant index (e.g., _main index, _test index). As such, each inverted index can correspond to a particular range of time for an index. Additional files, such as high performance indexes for each time-series bucket of an index, can also be stored in the same directory as the inverted indexes 507A . . . 507B, and 509A . . . 509B. In some embodiments, inverted index 507A . . . 507B, and 509A . . . 509B can correspond to multiple time-series buckets or inverted indexes 507A . . . 507B, and 509A . . . 509B can correspond to a single time-series bucket.
Each inverted index 507A . . . 507B, and 509A . . . 509B can include one or more entries, such as keyword (or token) entries or field-value pair entries. Furthermore, in certain embodiments, the inverted indexes 507A . . . 507B, and 509A . . . 509B can include additional information, such as a time range 523 associated with the inverted index or an index identifier 525 identifying the index associated with the inverted index 507A . . . 507B, and 509A . . . 509B. However, each inverted index 507A . . . 507B, and 509A . . . 509B can include less or more information than depicted.
Token entries, such as token entries 511 illustrated in inverted index 507B, can include a token 511A (e.g., “error,” “itemID,” etc.) and event references 511B indicative of events that include the token. For example, for the token “error,” the corresponding token entry includes the token “error” and an event reference, or unique identifier, for each event stored in the corresponding time-series bucket that includes the token “error.” In the illustrated embodiment of
In some cases, some token entries can be default entries, automatically determined entries, or user specified entries. In some embodiments, the indexer 206 can identify each word or string in an event as a distinct token and generate a token entry for it. In some cases, the indexer 206 can identify the beginning and ending of tokens based on punctuation, spaces, as described in greater detail herein. In certain cases, the indexer 206 can rely on user input or a configuration file to identify tokens for token entries 511, etc. It will be understood that any combination of token entries can be included as a default, automatically determined, a or included based on user-specified criteria.
Similarly, field-value pair entries, such as field-value pair entries 513 shown in inverted index 507B, can include a field-value pair 513A and event references 513B indicative of events that include a field value that corresponds to the field-value pair. For example, for a field-value pair sourcetype::sendmail, a field-value pair entry would include the field-value pair sourcetype::sendmail and a unique identifier, or event reference, for each event stored in the corresponding time-series bucket that includes a sendmail sourcetype.
In some cases, the field-value pair entries 513 can be default entries, automatically determined entries, or user specified entries. As a non-limiting example, the field-value pair entries for the fields host, source, sourcetype can be included in the inverted indexes 507A . . . 507B, and 509A . . . 509B as a default. As such, all of the inverted indexes 507A . . . 507B, and 509A . . . 509B can include field-value pair entries for the fields host, source, sourcetype. As yet another non-limiting example, the field-value pair entries for the IP_address field can be user specified and may only appear in the inverted index 507B based on user-specified criteria. As another non-limiting example, as the indexer indexes the events, it can automatically identify field-value pairs and create field-value pair entries. For example, based on the indexers review of events, it can identify IP_address as a field in each event and add the IP_address field-value pair entries to the inverted index 507B. It will be understood that any combination of field-value pair entries can be included as a default, automatically determined, or included based on user-specified criteria.
Each unique identifier 517, or event reference, can correspond to a unique event located in the time series bucket. However, the same event reference can be located in multiple entries. For example if an event has a sourcetype splunkd, host wwwl and token “warning,” then the unique identifier for the event will appear in the field-value pair entries sourcetype::splunkd and host::wwwl, as well as the token entry “warning.” With reference to the illustrated embodiment of
For some fields, the unique identifier is located in only one field-value pair entry for a particular field. For example, the inverted index may include four sourcetype field-value pair entries corresponding to four different sourcetypes of the events stored in a bucket (e.g., sourcetypes: sendmail, splunkd, web access, and web service). Within those four sourcetype field-value pair entries, an identifier for a particular event may appear in only one of the field-value pair entries. With continued reference to the example illustrated embodiment of
The event references 517 can be used to locate the events in the corresponding bucket. For example, the inverted index can include, or be associated with, an event reference array 515. The event reference array 515 can include an array entry 517 for each event reference in the inverted index 507B. Each array entry 517 can include location information 519 of the event corresponding to the unique identifier (non-limiting example: seek address of the event), a timestamp 521 associated with the event, or additional information regarding the event associated with the event reference, etc.
For each token entry 511 or field-value pair entry 513, the event reference 501Bor unique identifiers can be listed in chronological order or the value of the event reference can be assigned based on chronological data, such as a timestamp associated with the event referenced by the event reference. For example, the event reference 1 in the illustrated embodiment of
As a non-limiting example of how the inverted indexes 507A . . . 507B, and 509A . . . 509B can be used during a data categorization request command, the indexers can receive filter criteria indicating data that is to be categorized and categorization criteria indicating how the data is to be categorized. Example filter criteria can include, but is not limited to, indexes (or partitions), hosts, sources, sourcetypes, time ranges, field identifier, keywords, etc.
Using the filter criteria, the indexer identifies relevant inverted indexes to be searched. For example, if the filter criteria includes a set of partitions, the indexer can identify the inverted indexes stored in the directory corresponding to the particular partition as relevant inverted indexes. Other means can be used to identify inverted indexes associated with a partition of interest. For example, in some embodiments, the indexer can review an entry in the inverted indexes, such as an index-value pair entry 513 to determine if a particular inverted index is relevant. If the filter criteria does not identify any partition, then the indexer can identify all inverted indexes managed by the indexer as relevant inverted indexes.
Similarly, if the filter criteria includes a time range, the indexer can identify inverted indexes corresponding to buckets that satisfy at least a portion of the time range as relevant inverted indexes. For example, if the time range is last hour then the indexer can identify all inverted indexes that correspond to buckets storing events associated with timestamps within the last hour as relevant inverted indexes.
When used in combination, an index filter criterion specifying one or more partitions and a time range filter criterion specifying a particular time range can be used to identify a subset of inverted indexes within a particular directory (or otherwise associated with a particular partition) as relevant inverted indexes. As such, the indexer can focus the processing to only a subset of the total number of inverted indexes that the indexer manages.
Once the relevant inverted indexes are identified, the indexer can review them using any additional filter criteria to identify events that satisfy the filter criteria. In some cases, using the known location of the directory in which the relevant inverted indexes are located, the indexer can determine that any events identified using the relevant inverted indexes satisfy an index filter criterion. For example, if the filter criteria includes a partition main, then the indexer can determine that any events identified using inverted indexes within the partition main directory (or otherwise associated with the partition main) satisfy the index filter criterion.
Furthermore, based on the time range associated with each inverted index, the indexer can determine that that any events identified using a particular inverted index satisfies a time range filter criterion. For example, if a time range filter criterion is for the last hour and a particular inverted index corresponds to events within a time range of 50 minutes ago to 35 minutes ago, the indexer can determine that any events identified using the particular inverted index satisfy the time range filter criterion. Conversely, if the particular inverted index corresponds to events within a time range of 59 minutes ago to 62 minutes ago, the indexer can determine that some events identified using the particular inverted index may not satisfy the time range filter criterion.
Using the inverted indexes, the indexer can identify event references (and therefore events) that satisfy the filter criteria. For example, if the token “error” is a filter criterion, the indexer can track all event references within the token entry “error.” Similarly, the indexer can identify other event references located in other token entries or field-value pair entries that match the filter criteria. The system can identify event references located in all of the entries identified by the filter criteria. For example, if the filter criteria include the token “error” and field-value pair sourcetype::web_ui, the indexer can track the event references found in both the token entry “error” and the field-value pair entry sourcetype::web_ui. As mentioned previously, in some cases, such as when multiple values are identified for a particular filter criterion (e.g., multiple sources for a source filter criterion), the system can identify event references located in at least one of the entries corresponding to the multiple values and in all other entries identified by the filter criteria. The indexer can determine that the events associated with the identified event references satisfy the filter criteria.
In some cases, the indexer can further consult a timestamp associated with the event reference to determine whether an event satisfies the filter criteria. For example, if an inverted index corresponds to a time range that is partially outside of a time range filter criterion, then the indexer can consult a timestamp associated with the event reference to determine whether the corresponding event satisfies the time range criterion. In some embodiments, to identify events that satisfy a time range, the indexer can review an array, such as the event reference array1614 that identifies the time associated with the events. Furthermore, as mentioned above using the known location of the directory in which the relevant inverted indexes are located (or other index identifier), the indexer can determine that any events identified using the relevant inverted indexes satisfy the index filter criterion.
In some cases, based on the filter criteria, the indexer reviews an extraction rule. In certain embodiments, if the filter criteria includes a field name that does not correspond to a field-value pair entry in an inverted index, the indexer can review an extraction rule, which may be located in a configuration file, to identify a field that corresponds to a field-value pair entry in the inverted index.
For example, the filter criteria includes a field name “sessionID” and the indexer determines that at least one relevant inverted index does not include a field-value pair entry corresponding to the field name sessionID, the indexer can review an extraction rule that identifies how the sessionID field is to be extracted from a particular host, source, or sourcetype (implicitly identifying the particular host, source, or sourcetype that includes a sessionID field). The indexer can replace the field name “sessionID” in the filter criteria with the identified host, source, or sourcetype. In some cases, the field name “sessionID” may be associated with multiples hosts, sources, or sourcetypes, in which case, all identified hosts, sources, and sourcetypes can be added as filter criteria. In some cases, the identified host, source, or sourcetype can replace or be appended to a filter criterion, or be excluded. For example, if the filter criteria includes a criterion for source S1 and the “sessionID” field is found in source S2, the source S2 can replace S1 in the filter criteria, be appended such that the filter criteria includes source S1 and source S2, or be excluded based on the presence of the filter criterion source S1. If the identified host, source, or sourcetype is included in the filter criteria, the indexer can then identify a field-value pair entry in the inverted index that includes a field value corresponding to the identity of the particular host, source, or sourcetype identified using the extraction rule.
Once the events that satisfy the filter criteria are identified, the system, such as the indexer 206 can categorize the results based on the categorization criteria. The categorization criteria can include categories for grouping the results, such as any combination of partition, source, sourcetype, or host, or other categories or fields as desired.
The indexer can use the categorization criteria to identify categorization criteria-value pairs or categorization criteria values by which to categorize or group the results. The categorization criteria-value pairs can correspond to one or more field-value pair entries stored in a relevant inverted index, one or more index-value pairs based on a directory in which the inverted index is located or an entry in the inverted index (or other means by which an inverted index can be associated with a partition), or other criteria-value pair that identifies a general category and a particular value for that category. The categorization criteria values can correspond to the value portion of the categorization criteria-value pair.
As mentioned, in some cases, the categorization criteria-value pairs can correspond to one or more field-value pair entries stored in the relevant inverted indexes. For example, the categorization criteria-value pairs can correspond to field-value pair entries of host, source, and sourcetype (or other field-value pair entry as desired). For instance, if there are ten different hosts, four different sources, and five different sourcetypes for an inverted index, then the inverted index can include ten host field-value pair entries, four source field-value pair entries, and five sourcetype field-value pair entries. The indexer can use the nineteen distinct field-value pair entries as categorization criteria-value pairs to group the results.
Specifically, the indexer can identify the location of the event references associated with the events that satisfy the filter criteria within the field-value pairs, and group the event references based on their location. As such, the indexer can identify the particular field value associated with the event corresponding to the event reference. For example, if the categorization criteria include host and sourcetype, the host field-value pair entries and sourcetype field-value pair entries can be used as categorization criteria-value pairs to identify the specific host and sourcetype associated with the events that satisfy the filter criteria.
In addition, as mentioned, categorization criteria-value pairs can correspond to data other than the field-value pair entries in the relevant inverted indexes. For example, if partition or index is used as a categorization criterion, the inverted indexes may not include partition field-value pair entries. Rather, the indexer can identify the categorization criteria-value pair associated with the partition based on the directory in which an inverted index is located, information in the inverted index, or other information that associates the inverted index with the partition, etc. As such a variety of methods can be used to identify the categorization criteria-value pairs from the categorization criteria.
Accordingly based on the categorization criteria (and categorization criteria-value pairs), the indexer can generate groupings based on the events that satisfy the filter criteria. As a non-limiting example, if the categorization criteria includes a partition and sourcetype, then the groupings can correspond to events that are associated with each unique combination of partition and sourcetype. For instance, if there are three different partitions and two different sourcetypes associated with the identified events, then the six different groups can be formed, each with a unique partition value-sourcetype value combination. Similarly, if the categorization criteria includes partition, sourcetype, and host and there are two different partitions, three sourcetypes, and five hosts associated with the identified events, then the indexer can generate up to thirty groups for the results that satisfy the filter criteria. Each group can be associated with a unique combination of categorization criteria-value pairs (e.g., unique combinations of partition value sourcetype value, and host value).
In addition, the indexer can count the number of events associated with each group based on the number of events that meet the unique combination of categorization criteria for a particular group (or match the categorization criteria-value pairs for the particular group). With continued reference to the example above, the indexer can count the number of events that meet the unique combination of partition, sourcetype, and host for a particular group.
Each indexer communicates the groupings to the search head. The search head can aggregate the groupings from the indexers and provide the groupings for display. In some cases, the groups are displayed based on at least one of the host, source, sourcetype, or partition associated with the groupings. In some embodiments, the search head can further display the groups based on display criteria, such as a display order or a sort order as described in greater detail above.
As a non-limiting example and with reference to
Based on the above criteria, the indexer 206 identifies _main directory 503 and can ignore _test directory 505 and any other partition-specific directories. The indexer determines that inverted partition 507B is a relevant partition based on its location within the _main directory 503 and the time range associated with it. For sake of simplicity in this example, the indexer 206 determines that no other inverted indexes in the _main directory 503, such as inverted index 507A satisfy the time range criterion.
Having identified the relevant inverted index 507B, the indexer reviews the token entries 511 and the field-value pair entries 513 to identify event references, or events, that satisfy all of the filter criteria.
With respect to the token entries 511, the indexer can review the error token entry and identify event references 3, 5, 6, 8, 11, 12, indicating that the term “error” is found in the corresponding events. Similarly, the indexer can identify event references 4, 5, 6, 8, 9, 10, 11 in the field-value pair entry sourcetype::sourcetypeC and event references 2, 5, 6, 8, 10, 11 in the field-value pair entry host::hostB. As the filter criteria did not include a source or an IP_address field-value pair, the indexer can ignore those field-value pair entries.
In addition to identifying event references found in at least one token entry or field-value pair entry (e.g., event references 3, 4, 5, 6, 8, 9, 10, 11, 12), the indexer can identify events (and corresponding event references) that satisfy the time range criterion using the event reference array 1614 (e.g., event references 2, 3, 4, 5, 6, 7, 8, 9, 10). Using the information obtained from the inverted index 507B (including the event reference array 515), the indexer 206 can identify the event references that satisfy all of the filter criteria (e.g., event references 5, 6, 8).
Having identified the events (and event references) that satisfy all of the filter criteria, the indexer 206 can group the event references using the received categorization criteria (source). In doing so, the indexer can determine that event references 5 and 6 are located in the field-value pair entry source::sourceD (or have matching categorization criteria-value pairs) and event reference 8 is located in the field-value pair entry source::sourceC. Accordingly, the indexer can generate a sourceC group having a count of one corresponding to reference 8 and a sourceD group having a count of two corresponding to references 5 and 6. This information can be communicated to the search head. In turn the search head can aggregate the results from the various indexers and display the groupings. As mentioned above, in some embodiments, the groupings can be displayed based at least in part on the categorization criteria, including at least one of host, source, sourcetype, or partition.
It will be understood that a change to any of the filter criteria or categorization criteria can result in different groupings. As a one non-limiting example, a request received by an indexer 206 that includes the following filter criteria:partition=_main, time range=3/1/17 3/1/17 16:21:20.000-16:28:17.000, and the following categorization criteria:host, source, sourcetype would result in the indexer identifying event references 1-12 as satisfying the filter criteria. The indexer would then generate up to 24 groupings corresponding to the 24 different combinations of the categorization criteria-value pairs, including host (hostA, hostB), source (sourceA, sourceB, sourceC, sourceD), and sourcetype (sourcetypeA, sourcetypeB, sourcetypeC). However, as there are only twelve events identifiers in the illustrated embodiment and some fall into the same grouping, the indexer generates eight groups and counts as follows:
Group 1 (hostA, sourceA, sourcetypeA): 1 (event reference 7)
Group 2 (hostA, sourceA, sourcetypeB): 2 (event references 1, 12)
Group 3 (hostA, sourceA, sourcetypeC): 1 (event reference 4)
Group 4 (hostA, sourceB, sourcetypeA): 1 (event reference 3)
Group 5 (hostA, sourceB, sourcetypeC): 1 (event reference 9)
Group 6 (hostB, sourceC, sourcetypeA): 1 (event reference 2)
Group 7 (hostB, sourceC, sourcetypeC): 2 (event references 8, 11)
Group 8 (hostB, sourceD, sourcetypeC): 3 (event references 5, 6, 10)
As noted, each group has a unique combination of categorization criteria-value pairs or categorization criteria values. The indexer communicates the groups to the search head for aggregation with results received from other indexers. In communicating the groups to the search head, the indexer can include the categorization criteria-value pairs for each group and the count. In some embodiments, the indexer can include more or less information. For example, the indexer can include the event references associated with each group and other identifying information, such as the indexer or inverted index used to identify the groups.
As another non-limiting examples, a request received by an indexer 206 that includes the following filter criteria:partition=_main, time range=3/1/17 3/1/17 16:21:20.000-16:28:17.000, source=sourceA, sourceD, and keyword=itemID and the following categorization criteria:host, source, sourcetype would result in the indexer identifying event references 4, 7, and 10 as satisfying the filter criteria, and generate the following groups:
Group 1 (hostA, sourceA, sourcetypeC): 1 (event reference 4)
Group 2 (hostA, sourceA, sourcetypeA): 1 (event reference 7)
Group 3 (hostB, sourceD, sourcetypeC): 1 (event references 10)
The indexer communicates the groups to the search head for aggregation with results received from other indexers. As will be understand there are myriad ways for filtering and categorizing the events and event references. For example, the indexer can review multiple inverted indexes associated with an partition or review the inverted indexes of multiple partitions, and categorize the data using any one or any combination of partition, host, source, sourcetype, or other category, as desired.
Further, if a user interacts with a particular group, the indexer can provide additional information regarding the group. For example, the indexer can perform a targeted search or sampling of the events that satisfy the filter criteria and the categorization criteria for the selected group, also referred to as the filter criteria corresponding to the group or filter criteria associated with the group.
In some cases, to provide the additional information, the indexer relies on the inverted index. For example, the indexer can identify the event references associated with the events that satisfy the filter criteria and the categorization criteria for the selected group and then use the event reference array 515 to access some or all of the identified events. In some cases, the categorization criteria values or categorization criteria-value pairs associated with the group become part of the filter criteria for the review.
With reference to
In some embodiments, the indexer identifies the event references associated with the group using the filter criteria and the categorization criteria for the group (e.g., categorization criteria values or categorization criteria-value pairs unique to the group). Together, the filter criteria and the categorization criteria for the group can be referred to as the filter criteria associated with the group. Using the filter criteria associated with the group, the indexer identifies event references 4, 5, 6, 8, 10, 11.
Based on a sampling criteria, discussed in greater detail above, the indexer can determine that it will analyze a sample of the events associated with the event references 4, 5, 6, 8, 10, 11. For example, the sample can include analyzing event data associated with the event references 5, 8, 10. In some embodiments, the indexer can use the event reference array 1616 to access the event data associated with the event references 5, 8, 10. Once accessed, the indexer can compile the relevant information and provide it to the search head for aggregation with results from other indexers. By identifying events and sampling event data using the inverted indexes, the indexer can reduce the amount of actual data this is analyzed and the number of events that are accessed in order to generate the summary of the group and provide a response in less time.
2.8. Query Processing
At block 608, the indexers to which the query was distributed, search data stores associated with them for events that are responsive to the query. To determine which events are responsive to the query, the indexer searches for events that match the criteria specified in the query. These criteria can include matching keywords or specific values for certain fields. The searching operations at block 608 may use the late-binding schema to extract values for specified fields from events at the time the query is processed. In some embodiments, one or more rules for extracting field values may be specified as part of a source type definition in a configuration file. The indexers may then either send the relevant events back to the search head, or use the events to determine a partial result, and send the partial result back to the search head.
At block 610, the search head combines the partial results and/or events received from the indexers to produce a final result for the query. In some examples, the results of the query are indicative of performance or security of the IT environment and may help improve the performance of components in the IT environment. This final result may comprise different types of data depending on what the query requested. For example, the results can include a listing of matching events returned by the query, or some type of visualization of the data from the returned events. In another example, the final result can include one or more calculated values derived from the matching events.
The results generated by the system 108 can be returned to a client using different techniques. For example, one technique streams results or relevant events back to a client in real-time as they are identified. Another technique waits to report the results to the client until a complete set of results (which may include a set of relevant events or a result based on relevant events) is ready to return to the client. Yet another technique streams interim results or relevant events back to the client in real-time until a complete set of results is ready, and then returns the complete set of results to the client. In another technique, certain results are stored as “search jobs” and the client may retrieve the results by referring the search jobs.
The search head can also perform various operations to make the search more efficient. For example, before the search head begins execution of a query, the search head can determine a time range for the query and a set of common keywords that all matching events include. The search head may then use these parameters to query the indexers to obtain a superset of the eventual results. Then, during a filtering stage, the search head can perform field-extraction operations on the superset to produce a reduced set of search results. This speeds up queries, which may be particularly helpful for queries that are performed on a periodic basis.
2.9. Pipelined Search Language
Various embodiments of the present disclosure can be implemented using, or in conjunction with, a pipelined command language. A pipelined command language is a language in which a set of inputs or data is operated on by a first command in a sequence of commands, and then subsequent commands in the order they are arranged in the sequence. Such commands can include any type of functionality for operating on data, such as retrieving, searching, filtering, aggregating, processing, transmitting, and the like. As described herein, a query can thus be formulated in a pipelined command language and include any number of ordered or unordered commands for operating on data.
Splunk Processing Language (SPL) is an example of a pipelined command language in which a set of inputs or data is operated on by any number of commands in a particular sequence. A sequence of commands, or command sequence, can be formulated such that the order in which the commands are arranged defines the order in which the commands are applied to a set of data or the results of an earlier executed command. For example, a first command in a command sequence can operate to search or filter for specific data in particular set of data. The results of the first command can then be passed to another command listed later in the command sequence for further processing.
In various embodiments, a query can be formulated as a command sequence defined in a command line of a search UI. In some embodiments, a query can be formulated as a sequence of SPL commands. Some or all of the SPL commands in the sequence of SPL commands can be separated from one another by a pipe symbol “1”. In such embodiments, a set of data, such as a set of events, can be operated on by a first SPL command in the sequence, and then a subsequent SPL command following a pipe symbol “1” after the first SPL command operates on the results produced by the first SPL command or other set of data, and so on for any additional SPL commands in the sequence. As such, a query formulated using SPL comprises a series of consecutive commands that are delimited by pipe “1” characters. The pipe character indicates to the system that the output or result of one command (to the left of the pipe) should be used as the input for one of the subsequent commands (to the right of the pipe). This enables formulation of queries defined by a pipeline of sequenced commands that refines or enhances the data at each step along the pipeline until the desired results are attained. Accordingly, various embodiments described herein can be implemented with Splunk Processing Language (SPL) used in conjunction with the SPLUNK® ENTERPRISE system.
While a query can be formulated in many ways, a query can start with a search command and one or more corresponding search terms at the beginning of the pipeline. Such search terms can include any combination of keywords, phrases, times, dates, Boolean expressions, fieldname-field value pairs, etc. that specify which results should be obtained from an index. The results can then be passed as inputs into subsequent commands in a sequence of commands by using, for example, a pipe character. The subsequent commands in a sequence can include directives for additional processing of the results once it has been obtained from one or more indexes. For example, commands may be used to filter unwanted information out of the results, extract more information, evaluate field values, calculate statistics, reorder the results, create an alert, create summary of the results, or perform some type of aggregation function. In some embodiments, the summary can include a graph, chart, metric, or other visualization of the data. An aggregation function can include analysis or calculations to return an aggregate value, such as an average value, a sum, a maximum value, a root mean square, statistical values, and the like.
Due to its flexible nature, use of a pipelined command language in various embodiments is advantageous because it can perform “filtering” as well as “processing” functions. In other words, a single query can include a search command and search term expressions, as well as data-analysis expressions. For example, a command at the beginning of a query can perform a “filtering” step by retrieving a set of data based on a condition (e.g., records associated with server response times of less than 1 microsecond). The results of the filtering step can then be passed to a subsequent command in the pipeline that performs a “processing” step (e.g. calculation of an aggregate value related to the filtered events such as the average response time of servers with response times of less than 1 microsecond). Furthermore, the search command can allow events to be filtered by keyword as well as field value criteria. For example, a search command can filter out all events containing the word “warning” or filter out all events where a field value associated with a field “clientip” is “10.0.1.2.”
The results obtained or generated in response to a command in a query can be considered a set of results data. The set of results data can be passed from one command to another in any data format. In one embodiment, the set of result data can be in the form of a dynamically created table. Each command in a particular query can redefine the shape of the table. In some implementations, an event retrieved from an index in response to a query can be considered a row with a column for each field value. Columns contain basic information about the data and also may contain data that has been dynamically extracted at search time.
Disk 622 represents the event data in the raw record data store.
When a user query is processed, a search step will precede other queries in the pipeline in order to generate a set of events at block 640. For example, the query can comprise search terms “sourcetype=syslog ERROR” at the front of the pipeline as shown in
At block 642, the set of events generated in the first part of the query may be piped to a query that searches the set of events for field-value pairs or for keywords. For example, the second intermediate results table 626 shows fewer columns, representing the result of the top command, “top user” which summarizes the events into a list of the top 10 users and displays the user, count, and percentage.
Finally, at block 644, the results of the prior stage can be pipelined to another stage where further filtering or processing of the data can be performed, e.g., preparing the data for display purposes, filtering the data based on a condition, performing a mathematical calculation with the data, etc. As shown in
2.10. Field Extraction
The search head 210 allows users to search and visualize events generated from machine data received from homogenous data sources. The search head 210 also allows users to search and visualize events generated from machine data received from heterogeneous data sources. The search head 210 includes various mechanisms, which may additionally reside in an indexer 206, for processing a query. A query language may be used to create a query, such as any suitable pipelined query language. For example, Splunk Processing Language (SPL) can be utilized to make a query. SPL is a pipelined search language in which a set of inputs is operated on by a first command in a command line, and then a subsequent command following the pipe symbol “1” operates on the results produced by the first command, and so on for additional commands. Other query languages, such as the Structured Query Language (“SQL”), can be used to create a query.
In response to receiving the search query, search head 210 uses extraction rules to extract values for fields in the events being searched. The search head 210 obtains extraction rules that specify how to extract a value for fields from an event. Extraction rules can comprise regex rules that specify how to extract values for the fields corresponding to the extraction rules. In addition to specifying how to extract field values, the extraction rules may also include instructions for deriving a field value by performing a function on a character string or value retrieved by the extraction rule. For example, an extraction rule may truncate a character string or convert the character string into a different data format. In some cases, the query itself can specify one or more extraction rules.
The search head 210 can apply the extraction rules to events that it receives from indexers 206. Indexers 206 may apply the extraction rules to events in an associated data store 208. Extraction rules can be applied to all the events in a data store or to a subset of the events that have been filtered based on some criteria (e.g., event time stamp values, etc.). Extraction rules can be used to extract one or more values for a field from events by parsing the portions of machine data in the events and examining the data for one or more patterns of characters, numbers, delimiters, etc., that indicate where the field begins and, optionally, ends.
Using the log data received at one or more indexers 206 from the three systems, the vendor can uniquely obtain an insight into user activity, user experience, and system behavior. The search head 210 allows the vendor's administrator to search the log data from the three systems that one or more indexers 206 are responsible for searching, thereby obtaining correlated information, such as the order number and corresponding customer ID number of the person placing the order. The system also allows the administrator to see a visualization of related events via a user interface. The administrator can query the search head 210 for customer ID field value matches across the log data from the three systems that are stored at the one or more indexers 206. The customer ID field value exists in the data gathered from the three systems, but the customer ID field value may be located in different areas of the data given differences in the architecture of the systems. There is a semantic relationship between the customer ID field values generated by the three systems. The search head 210 requests events from the one or more indexers 206 to gather relevant events from the three systems. The search head 210 then applies extraction rules to the events in order to extract field values that it can correlate. The search head may apply a different extraction rule to each set of events from each system when the event format differs among systems. In this example, the user interface can display to the administrator the events corresponding to the common customer ID field values 707, 708, and 709, thereby providing the administrator with insight into a customer's experience.
Note that query results can be returned to a client, a search head, or any other system component for further processing. In general, query results may include a set of one or more events, a set of one or more values obtained from the events, a subset of the values, statistics calculated based on the values, a report containing the values, a visualization (e.g., a graph or chart) generated from the values, and the like.
The search system enables users to run queries against the stored data to retrieve events that meet criteria specified in a query, such as containing certain keywords or having specific values in defined fields.
If a user inputs a search query into search bar 1401 that includes only keywords (also known as “tokens”), e.g., the keyword “error” or “warning”, the query search engine of the data intake and query system searches for those keywords directly in the event data 722 stored in the raw record data store. Note that while
As disclosed above, an indexer can optionally generate a keyword index to facilitate fast keyword searching for event data. The indexer includes the identified keywords in an index, which associates each stored keyword with reference pointers to events containing that keyword (or to locations within events where that keyword is located, other location identifiers, etc.). When an indexer subsequently receives a keyword-based query, the indexer can access the keyword index to quickly identify events containing the keyword. For example, if the keyword “HTTP” was indexed by the indexer at index time, and the user searches for the keyword “HTTP”, events 713 to 715 will be identified based on the results returned from the keyword index. As noted above, the index contains reference pointers to the events containing the keyword, which allows for efficient retrieval of the relevant events from the raw record data store.
If a user searches for a keyword that has not been indexed by the indexer, the data intake and query system would nevertheless be able to retrieve the events by searching the event data for the keyword in the raw record data store directly as shown in
In most cases, however, in addition to keywords, a user's search will also include fields. The term “field” refers to a location in the event data containing one or more values for a specific data item. Often, a field is a value with a fixed, delimited position on a line, or a name and value pair, where there is a single value to each field name. A field can also be multivalued, that is, it can appear more than once in an event and have a different value for each appearance, e.g., email address fields. Fields are searchable by the field name or field name-value pairs. Some examples of fields are “clientip” for IP addresses accessing a web server, or the “From” and “To” fields in email addresses.
By way of further example, consider the search, “status=404”. This search query finds events with “status” fields that have a value of “404.” When the search is run, the search engine does not look for events with any other “status” value. It also does not look for events containing other fields that share “404” as a value. As a result, the search returns a set of results that are more focused than if “404” had been used in the search string as part of a keyword search. Note also that fields can appear in events as “key=value” pairs such as “user_name=Bob.” But in most cases, field values appear in fixed, delimited positions without identifying keys. For example, the data store may contain events where the “user_name” value always appears by itself after the timestamp as illustrated by the following string: “Nov. 15 09:33:22 johnmedlock.”
The data intake and query system advantageously allows for search time field extraction. In other words, fields can be extracted from the event data at search time using late-binding schema as opposed to at data ingestion time, which was a major limitation of the prior art systems.
In response to receiving the search query, search head 210 uses extraction rules to extract values for the fields associated with a field or fields in the event data being searched. The search head 210 obtains extraction rules that specify how to extract a value for certain fields from an event. Extraction rules can comprise regex rules that specify how to extract values for the relevant fields. In addition to specifying how to extract field values, the extraction rules may also include instructions for deriving a field value by performing a function on a character string or value retrieved by the extraction rule. For example, a transformation rule may truncate a character string, or convert the character string into a different data format. In some cases, the query itself can specify one or more extraction rules.
Configuration file 712 may contain extraction rules for all the various fields that are not metadata fields, e.g., the “clientip” field. The extraction rules may be inserted into the configuration file in a variety of ways. In some embodiments, the extraction rules can comprise regular expression rules that are manually entered in by the user. Regular expressions match patterns of characters in text and are used for extracting custom fields in text.
In one or more embodiments, as noted above, a field extractor may be configured to automatically generate extraction rules for certain field values in the events when the events are being created, indexed, or stored, or possibly at a later time. In one embodiment, a user may be able to dynamically create custom fields by highlighting portions of a sample event that should be extracted as fields using a graphical user interface. The system would then generate a regular expression that extracts those fields from similar events and store the regular expression as an extraction rule for the associated field in the configuration file 712.
In some embodiments, the indexers may automatically discover certain custom fields at index time and the regular expressions for those fields will be automatically generated at index time and stored as part of extraction rules in configuration file 712. For example, fields that appear in the event data as “key=value” pairs may be automatically extracted as part of an automatic field discovery process. Note that there may be several other ways of adding field definitions to configuration files in addition to the methods discussed herein.
The search head 210 can apply the extraction rules derived from configuration file 1402 to event data that it receives from indexers 206. Indexers 206 may apply the extraction rules from the configuration file to events in an associated data store 208. Extraction rules can be applied to all the events in a data store, or to a subset of the events that have been filtered based on some criteria (e.g., event time stamp values, etc.). Extraction rules can be used to extract one or more values for a field from events by parsing the event data and examining the event data for one or more patterns of characters, numbers, delimiters, etc., that indicate where the field begins and, optionally, ends.
In one more embodiments, the extraction rule in configuration file 712 will also need to define the type or set of events that the rule applies to. Because the raw record data store will contain events from multiple heterogeneous sources, multiple events may contain the same fields in different locations because of discrepancies in the format of the data generated by the various sources. Furthermore, certain events may not contain a particular field at all. For example, event 719 also contains “clientip” field, however, the “clientip” field is in a different format from events 713-715. To address the discrepancies in the format and content of the different types of events, the configuration file will also need to specify the set of events that an extraction rule applies to, e.g., extraction rule 716 specifies a rule for filtering by the type of event and contains a regular expression for parsing out the field value. Accordingly, each extraction rule will pertain to only a particular type of event. If a particular field, e.g., “clientip” occurs in multiple events, each of those types of events would need its own corresponding extraction rule in the configuration file 712 and each of the extraction rules would comprise a different regular expression to parse out the associated field value. The most common way to categorize events is by source type because events generated by a particular source can have the same format.
The field extraction rules stored in configuration file 712 perform search-time field extractions. For example, for a query that requests a list of events with source type “access_combined” where the “clientip” field equals “127.0.0.1,” the query search engine would first locate the configuration file 712 to retrieve extraction rule 716 that would allow it to extract values associated with the “clientip” field from the event data 720 “where the source type is “access_combined. After the “clientip” field has been extracted from all the events comprising the “clientip” field where the source type is “access_combined,” the query search engine can then execute the field criteria by performing the compare operation to filter out the events where the “clientip” field equals “127.0.0.1.” In the example shown in
The configuration file can be created during indexing. It may either be manually created by the user or automatically generated with certain predetermined field extraction rules. As discussed above, the events may be distributed across several indexers, wherein each indexer may be responsible for storing and searching a subset of the events contained in a corresponding data store. In a distributed indexer system, each indexer would need to maintain a local copy of the configuration file that is synchronized periodically across the various indexers.
The ability to add schema to the configuration file at search time results in increased efficiency. A user can create new fields at search time and simply add field definitions to the configuration file. As a user learns more about the data in the events, the user can continue to refine the late-binding schema by adding new fields, deleting fields, or modifying the field extraction rules in the configuration file for use the next time the schema is used by the system. Because the data intake and query system maintains the underlying raw data and uses late-binding schema for searching the raw data, it enables a user to continue investigating and learn valuable insights about the raw data long after data ingestion time.
The ability to add multiple field definitions to the configuration file at search time also results in increased flexibility. For example, multiple field definitions can be added to the configuration file to capture the same field across events generated by different source types. This allows the data intake and query system to search and correlate data across heterogeneous sources flexibly and efficiently.
Further, by providing the field definitions for the queried fields at search time, the configuration file 712 allows the record data store 712 to be field searchable. In other words, the raw record data store 712 can be searched using keywords as well as fields, wherein the fields are searchable name/value pairings that distinguish one event from another and can be defined in configuration file 1402 using extraction rules. In comparison to a search containing field names, a keyword search does not need the configuration file and can search the event data directly as shown in
It should also be noted that any events filtered out by performing a search-time field extraction using a configuration file can be further processed by directing the results of the filtering step to a processing step using a pipelined search language. Using the prior example, a user could pipeline the results of the compare step to an aggregate function by asking the query search engine to count the number of events where the “clientip” field equals “127.0.0.1.”
2.11. Example Search Screen
After the search is executed, the search screen 800 in
The events tab additionally displays a sidebar that is an interactive field picker 806. The field picker 806 may be displayed to a user in response to the search being executed and allows the user to further analyze the search results based on the fields in the events of the search results. The field picker 806 includes field names that reference fields present in the events in the search results. The field picker may display any Selected Fields 820 that a user has pre-selected for display (e.g., host, source, sourcetype) and may also display any Interesting Fields 822 that the system determines may be interesting to the user based on pre-specified criteria (e.g., action, bytes, categoryid, clientip, date_hour, date_mday, date_minute, etc.). The field picker also provides an option to display field names for all the fields present in the events of the search results using the All Fields control 824.
Each field name in the field picker 806 has a value type identifier to the left of the field name, such as value type identifier 826. A value type identifier identifies the type of value for the respective field, such as an “a” for fields that include literal values or a “#” for fields that include numerical values.
Each field name in the field picker also has a unique value count to the right of the field name, such as unique value count 828. The unique value count indicates the number of unique values for the respective field in the events of the search results.
Each field name is selectable to view the events in the search results that have the field referenced by that field name. For example, a user can select the “host” field name, and the events shown in the events list 808 will be updated with events in the search results that have the field that is reference by the field name “host.”
2.12. Data Models
A data model is a hierarchically structured search-time mapping of semantic knowledge about one or more datasets. It encodes the domain knowledge used to build a variety of specialized searches of those datasets. Those searches, in turn, can be used to generate reports.
A data model is composed of one or more “objects” (or “data model objects”) that define or otherwise correspond to a specific set of data. An object is defined by constraints and attributes. An object's constraints are search criteria that define the set of events to be operated on by running a search having that search criteria at the time the data model is selected. An object's attributes are the set of fields to be exposed for operating on that set of events generated by the search criteria.
Objects in data models can be arranged hierarchically in parent/child relationships. Each child object represents a subset of the dataset covered by its parent object. The top-level objects in data models are collectively referred to as “root objects.”
Child objects have inheritance. Child objects inherit constraints and attributes from their parent objects and may have additional constraints and attributes of their own. Child objects provide a way of filtering events from parent objects. Because a child object may provide an additional constraint in addition to the constraints it has inherited from its parent object, the dataset it represents may be a subset of the dataset that its parent represents. For example, a first data model object may define a broad set of data pertaining to e-mail activity generally, and another data model object may define specific datasets within the broad dataset, such as a subset of the e-mail data pertaining specifically to e-mails sent. For example, a user can simply select an “e-mail activity” data model object to access a dataset relating to e-mails generally (e.g., sent or received), or select an “e-mails sent” data model object (or data sub-model object) to access a dataset relating to e-mails sent.
Because a data model object is defined by its constraints (e.g., a set of search criteria) and attributes (e.g., a set of fields), a data model object can be used to quickly search data to identify a set of events and to identify a set of fields to be associated with the set of events. For example, an “e-mails sent” data model object may specify a search for events relating to e-mails that have been sent, and specify a set of fields that are associated with the events. Thus, a user can retrieve and use the “e-mails sent” data model object to quickly search source data for events relating to sent e-mails, and may be provided with a listing of the set of fields relevant to the events in a user interface screen.
Examples of data models can include electronic mail, authentication, databases, intrusion detection, malware, application state, alerts, compute inventory, network sessions, network traffic, performance, audits, updates, vulnerabilities, etc. Data models and their objects can be designed by knowledge managers in an organization, and they can enable downstream users to quickly focus on a specific set of data. A user iteratively applies a model development tool (not shown in
Data definitions in associated schemas can be taken from the common information model (CIM) or can be devised for a particular schema and optionally added to the CIM. Child objects inherit fields from parents and can include fields not present in parents. A model developer can select fewer extraction rules than are available for the sources returned by the query that defines events belonging to a model. Selecting a limited set of extraction rules can be a tool for simplifying and focusing the data model, while allowing a user flexibility to explore the data subset. Development of a data model is further explained in U.S. Pat. Nos. 8,788,525 and 8,788,526, both entitled “DATA MODEL FOR MACHINE DATA FOR SEMANTIC SEARCH”, both issued on 22 Jul. 2014, U.S. Pat. No. 8,983,994, entitled “GENERATION OF A DATA MODEL FOR SEARCHING MACHINE DATA”, issued on 17 Mar. 2015, U.S. Pat. No. 9,128,980, entitled “GENERATION OF A DATA MODEL APPLIED TO QUERIES”, issued on 8 Sep. 2015, and U.S. Pat. No. 9,589,012, entitled “GENERATION OF A DATA MODEL APPLIED TO OBJECT QUERIES”, issued on 7 Mar. 2017, each of which is hereby incorporated by reference in its entirety for all purposes.
A data model can also include reports. One or more report formats can be associated with a particular data model and be made available to run against the data model. A user can use child objects to design reports with object datasets that already have extraneous data pre-filtered out. In some embodiments, the data intake and query system 108 provides the user with the ability to produce reports (e.g., a table, chart, visualization, etc.) without having to enter SPL, SQL, or other query language terms into a search screen. Data models are used as the basis for the search feature.
Data models may be selected in a report generation interface. The report generator supports drag-and-drop organization of fields to be summarized in a report. When a model is selected, the fields with available extraction rules are made available for use in the report. The user may refine and/or filter search results to produce more precise reports. The user may select some fields for organizing the report and select other fields for providing detail according to the report organization. For example, “region” and “salesperson” are fields used for organizing the report and sales data can be summarized (subtotaled and totaled) within this organization. The report generator allows the user to specify one or more fields within events and apply statistical analysis on values extracted from the specified one or more fields. The report generator may aggregate search results across sets of events and generate statistics based on aggregated search results. Building reports using the report generation interface is further explained in U.S. patent application Ser. No. 14/503,335, entitled “GENERATING REPORTS FROM UNSTRUCTURED DATA”, filed on 30 Sep. 2014, and which is hereby incorporated by reference in its entirety for all purposes. Data visualizations also can be generated in a variety of formats, by reference to the data model. Reports, data visualizations, and data model objects can be saved and associated with the data model for future use. The data model object may be used to perform searches of other data.
The selection of a data model object for use in driving a report generation may be facilitated by a data model object selection interface.
Once a data model object is selected by the user, a user interface screen 1100 shown in
The reporting application allows the user to create graphical visualizations of the statistics generated for a report. For example,
2.13. Acceleration Technique
The above-described system provides significant flexibility by enabling a user to analyze massive quantities of minimally-processed data “on the fly” at search time using a late-binding schema, instead of storing pre-specified portions of the data in a database at ingestion time. This flexibility enables a user to see valuable insights, correlate data, and perform subsequent queries to examine interesting aspects of the data that may not have been apparent at ingestion time.
However, performing extraction and analysis operations at search time can involve a large amount of data and require a large number of computational operations, which can cause delays in processing the queries. Advantageously, the data intake and query system also employs a number of unique acceleration techniques that have been developed to speed up analysis operations performed at search time. These techniques include: (1) performing search operations in parallel across multiple indexers; (2) using a keyword index; (3) using a high performance analytics store; and (4) accelerating the process of generating reports. These novel techniques are described in more detail below.
2.13.1. Aggregation Technique
To facilitate faster query processing, a query can be structured such that multiple indexers perform the query in parallel, while aggregation of search results from the multiple indexers is performed locally at the search head. For example,
During operation, upon receiving search query 1602, a search head 210 determines that a portion of the operations involved with the search query may be performed locally by the search head. The search head modifies search query 1602 by substituting “stats” (create aggregate statistics over results sets received from the indexers at the search head) with “prestats” (create statistics by the indexer from local results set) to produce search query 1604, and then distributes search query 1604 to distributed indexers, which are also referred to as “search peers” or “peer indexers.” Note that search queries may generally specify search criteria or operations to be performed on events that meet the search criteria. Search queries may also specify field names, as well as search criteria for the values in the fields or operations to be performed on the values in the fields. Moreover, the search head may distribute the full search query to the search peers as illustrated in
2.13.2. Keyword Index
As described above with reference to the flow diagrams in
2.13.3. High Performance Analytics Store
To speed up certain types of queries, some embodiments of system 108 create a high performance analytics store, which is referred to as a “summarization table,” that contains entries for specific field-value pairs. Each of these entries keeps track of instances of a specific value in a specific field in the events and includes references to events containing the specific value in the specific field. For example, an example entry in a summarization table can keep track of occurrences of the value “94107” in a “ZIP code” field of a set of events and the entry includes references to all of the events that contain the value “94107” in the ZIP code field. This optimization technique enables the system to quickly process queries that seek to determine how many events have a particular value for a particular field. To this end, the system can examine the entry in the summarization table to count instances of the specific value in the field without having to go through the individual events or perform data extractions at search time. Also, if the system needs to process all events that have a specific field-value combination, the system can use the references in the summarization table entry to directly access the events to extract further information without having to search all of the events to find the specific field-value combination at search time.
In some embodiments, the system maintains a separate summarization table for each of the above-described time-specific buckets that stores events for a specific time range. A bucket-specific summarization table includes entries for specific field-value combinations that occur in events in the specific bucket. Alternatively, the system can maintain a separate summarization table for each indexer. The indexer-specific summarization table includes entries for the events in a data store that are managed by the specific indexer. Indexer-specific summarization tables may also be bucket-specific.
The summarization table can be populated by running a periodic query that scans a set of events to find instances of a specific field-value combination, or alternatively instances of all field-value combinations for a specific field. A periodic query can be initiated by a user, or can be scheduled to occur automatically at specific time intervals. A periodic query can also be automatically launched in response to a query that asks for a specific field-value combination.
In some cases, when the summarization tables may not cover all of the events that are relevant to a query, the system can use the summarization tables to obtain partial results for the events that are covered by summarization tables, but may also have to search through other events that are not covered by the summarization tables to produce additional results. These additional results can then be combined with the partial results to produce a final set of results for the query. The summarization table and associated techniques are described in more detail in U.S. Pat. No. 8,682,925, entitled “Distributed High Performance Analytics Store”, issued on 25 Mar. 2014, U.S. Pat. No. 9,128,985, entitled “SUPPLEMENTING A HIGH PERFORMANCE ANALYTICS STORE WITH EVALUATION OF INDIVIDUAL EVENTS TO RESPOND TO AN EVENT QUERY”, issued on 8 Sep. 2015, and U.S. patent application Ser. No. 14/815,973, entitled “GENERATING AND STORING SUMMARIZATION TABLES FOR SETS OF SEARCHABLE EVENTS”, filed on 1 Aug. 2015, each of which is hereby incorporated by reference in its entirety for all purposes.
To speed up certain types of queries, e.g., frequently encountered queries or computationally intensive queries, some embodiments of system 108 create a high performance analytics store, which is referred to as a “summarization table,” (also referred to as a “lexicon” or “inverted index”) that contains entries for specific field-value pairs. Each of these entries keeps track of instances of a specific value in a specific field in the event data and includes references to events containing the specific value in the specific field. For example, an example entry in an inverted index can keep track of occurrences of the value “94107” in a “ZIP code” field of a set of events and the entry includes references to all of the events that contain the value “94107” in the ZIP code field. Creating the inverted index data structure avoids needing to incur the computational overhead each time a statistical query needs to be run on a frequently encountered field-value pair. In order to expedite queries, in most embodiments, the search engine will employ the inverted index separate from the raw record data store to generate responses to the received queries.
Note that the term “summarization table” or “inverted index” as used herein is a data structure that may be generated by an indexer that includes at least field names and field values that have been extracted and/or indexed from event records. An inverted index may also include reference values that point to the location(s) in the field searchable data store where the event records that include the field may be found. Also, an inverted index may be stored using well-known compression techniques to reduce its storage size.
Further, note that the term “reference value” (also referred to as a “posting value”) as used herein is a value that references the location of a source record in the field searchable data store. In some embodiments, the reference value may include additional information about each record, such as timestamps, record size, meta-data, or the like. Each reference value may be a unique identifier which may be used to access the event data directly in the field searchable data store. In some embodiments, the reference values may be ordered based on each event record's timestamp. For example, if numbers are used as identifiers, they may be sorted so event records having a later timestamp always have a lower valued identifier than event records with an earlier timestamp, or vice-versa. Reference values are often included in inverted indexes for retrieving and/or identifying event records.
In one or more embodiments, an inverted index is generated in response to a user-initiated collection query. The term “collection query” as used herein refers to queries that include commands that generate summarization information and inverted indexes (or summarization tables) from event records stored in the field searchable data store.
Note that a collection query is a special type of query that can be user-generated and is used to create an inverted index. A collection query is not the same as a query that is used to call up or invoke a pre-existing inverted index. In one or more embodiment, a query can comprise an initial step that calls up a pre-generated inverted index on which further filtering and processing can be performed. For example, referring back to
In one or more embodiments, if one or more of the queries is a collection query, the responsive indexers may generate summarization information based on the fields of the event records located in the field searchable data store. In at least one of the various embodiments, one or more of the fields used in the summarization information may be listed in the collection query and/or they may be determined based on terms included in the collection query. For example, a collection query may include an explicit list of fields to summarize. Or, in at least one of the various embodiments, a collection query may include terms or expressions that explicitly define the fields, e.g., using regex rules. In
In one or more embodiments, collection queries may be saved and scheduled to run periodically. These scheduled collection queries may periodically update the summarization information corresponding to the query. For example, if the collection query that generates inverted index 722 is scheduled to run periodically, one or more indexers would periodically search through the relevant buckets to update inverted index 722 with event data for any new events with the “clientip” value of “127.0.0.1.”
In some embodiments, the inverted indexes that include fields, values, and reference value (e.g., inverted index 722) for event records may be included in the summarization information provided to the user. In other embodiments, a user may not be interested in specific fields and values contained in the inverted index, but may need to perform a statistical query on the data in the inverted index. For example, referencing the example of
The pipelined search language, e.g., SPL of the SPLUNK® ENTERPRISE system can be used to pipe the contents of an inverted index to a statistical query using the “stats” command for example. A “stats” query refers to queries that generate result sets that may produce aggregate and statistical results from event records, e.g., average, mean, max, min, rms, etc. Where sufficient information is available in an inverted index, a “stats” query may generate their result sets rapidly from the summarization information available in the inverted index rather than directly scanning event records. For example, the contents of inverted index 722 can be pipelined to a stats query, e.g., a “count” function that counts the number of entries in the inverted index and returns a value of “4.” In this way, inverted indexes may enable various stats queries to be performed absent scanning or search the event records. Accordingly, this optimization technique enables the system to quickly process queries that seek to determine how many events have a particular value for a particular field. To this end, the system can examine the entry in the inverted index to count instances of the specific value in the field without having to go through the individual events or perform data extractions at search time.
In some embodiments, the system maintains a separate inverted index for each of the above-described time-specific buckets that stores events for a specific time range. A bucket-specific inverted index includes entries for specific field-value combinations that occur in events in the specific bucket. Alternatively, the system can maintain a separate inverted index for each indexer. The indexer-specific inverted index includes entries for the events in a data store that are managed by the specific indexer. Indexer-specific inverted indexes may also be bucket-specific. In at least one or more embodiments, if one or more of the queries is a stats query, each indexer may generate a partial result set from previously generated summarization information. The partial result sets may be returned to the search head that received the query and combined into a single result set for the query
As mentioned above, the inverted index can be populated by running a periodic query that scans a set of events to find instances of a specific field-value combination, or alternatively instances of all field-value combinations for a specific field. A periodic query can be initiated by a user, or can be scheduled to occur automatically at specific time intervals. A periodic query can also be automatically launched in response to a query that asks for a specific field-value combination. In some embodiments, if summarization information is absent from an indexer that includes responsive event records, further actions may be taken, such as, the summarization information may generated on the fly, warnings may be provided the user, the collection query operation may be halted, the absence of summarization information may be ignored, or the like, or combination thereof.
In one or more embodiments, an inverted index may be set up to update continually. For example, the query may ask for the inverted index to update its result periodically, e.g., every hour. In such instances, the inverted index may be a dynamic data structure that is regularly updated to include information regarding incoming events.
In some cases, e.g., where a query is executed before an inverted index updates, when the inverted index may not cover all of the events that are relevant to a query, the system can use the inverted index to obtain partial results for the events that are covered by inverted index, but may also have to search through other events that are not covered by the inverted index to produce additional results on the fly. In other words, an indexer would need to search through event data on the data store to supplement the partial results. These additional results can then be combined with the partial results to produce a final set of results for the query. Note that in typical instances where an inverted index is not completely up to date, the number of events that an indexer would need to search through to supplement the results from the inverted index would be relatively small. In other words, the search to get the most recent results can be quick and efficient because only a small number of event records will be searched through to supplement the information from the inverted index. The inverted index and associated techniques are described in more detail in U.S. Pat. No. 8,682,925, entitled “Distributed High Performance Analytics Store”, issued on 25 Mar. 2014, U.S. Pat. No. 9,128,985, entitled “SUPPLEMENTING A HIGH PERFORMANCE ANALYTICS STORE WITH EVALUATION OF INDIVIDUAL EVENTS TO RESPOND TO AN EVENT QUERY”, filed on 31 Jan. 2014, and U.S. patent application Ser. No. 14/815,973, entitled “STORAGE MEDIUM AND CONTROL DEVICE”, filed on 21 Feb. 2014, each of which is hereby incorporated by reference in its entirety.
2.13.3.1 Extracting Event Data Using Posting
In one or more embodiments, if the system needs to process all events that have a specific field-value combination, the system can use the references in the inverted index entry to directly access the events to extract further information without having to search all of the events to find the specific field-value combination at search time. In other words, the system can use the reference values to locate the associated event data in the field searchable data store and extract further information from those events, e.g., extract further field values from the events for purposes of filtering or processing or both.
The information extracted from the event data using the reference values can be directed for further filtering or processing in a query using the pipeline search language. The pipelined search language will, in one embodiment, include syntax that can direct the initial filtering step in a query to an inverted index. In one embodiment, a user would include syntax in the query that explicitly directs the initial searching or filtering step to the inverted index.
Referencing the example in
In one embodiment, the same methodology can be used to pipe the contents of the inverted index to a processing step. In other words, the user is able to use the inverted index to efficiently and quickly perform aggregate functions on field values that were not part of the initially generated inverted index. For example, a user may want to determine an average object size (size of the requested gif) requested by clients from IP address “127.0.0.1.” In this case, the search engine would again use the reference values stored in inverted index 722 to retrieve the event data from the field searchable data store and, further, extract the object size field values from the associated events 731, 732, 733 and 734. Once, the corresponding object sizes have been extracted (i.e. 2326, 2900, 2920, and 5000), the average can be computed and returned to the user.
In one embodiment, instead of explicitly invoking the inverted index in a user-generated query, e.g., by the use of special commands or syntax, the SPLUNK® ENTERPRISE system can be configured to automatically determine if any prior-generated inverted index can be used to expedite a user query. For example, the user's query may request the average object size (size of the requested gif) requested by clients from IP address “127.0.0.1.” without any reference to or use of inverted index 722. The search engine, in this case, would automatically determine that an inverted index 722 already exists in the system that could expedite this query. In one embodiment, prior to running any search comprising a field-value pair, for example, a search engine may search though all the existing inverted indexes to determine if a pre-generated inverted index could be used to expedite the search comprising the field-value pair. Accordingly, the search engine would automatically use the pre-generated inverted index, e.g., index 722 to generate the results without any user-involvement that directs the use of the index.
Using the reference values in an inverted index to be able to directly access the event data in the field searchable data store and extract further information from the associated event data for further filtering and processing is highly advantageous because it avoids incurring the computation overhead of regenerating the inverted index with additional fields or performing a new search.
The data intake and query system includes one or more forwarders that receive raw machine data from a variety of input data sources, and one or more indexers that process and store the data in one or more data stores. By distributing events among the indexers and data stores, the indexers can analyze events for a query in parallel. In one or more embodiments, a multiple indexer implementation of the search system would maintain a separate and respective inverted index for each of the above-described time-specific buckets that stores events for a specific time range. A bucket-specific inverted index includes entries for specific field-value combinations that occur in events in the specific bucket. As explained above, a search head would be able to correlate and synthesize data from across the various buckets and indexers.
This feature advantageously expedites searches because instead of performing a computationally intensive search in a centrally located inverted index that catalogues all the relevant events, an indexer is able to directly search an inverted index stored in a bucket associated with the time-range specified in the query. This allows the search to be performed in parallel across the various indexers. Further, if the query requests further filtering or processing to be conducted on the event data referenced by the locally stored bucket-specific inverted index, the indexer is able to simply access the event records stored in the associated bucket for further filtering and processing instead of needing to access a central repository of event records, which would dramatically add to the computational overhead.
In one embodiment, there may be multiple buckets associated with the time-range specified in a query. If the query is directed to an inverted index, or if the search engine automatically determines that using an inverted index would expedite the processing of the query, the indexers will search through each of the inverted indexes associated with the buckets for the specified time-range. This feature allows the High Performance Analytics Store to be scaled easily.
In certain instances, where a query is executed before a bucket-specific inverted index updates, when the bucket-specific inverted index may not cover all of the events that are relevant to a query, the system can use the bucket-specific inverted index to obtain partial results for the events that are covered by bucket-specific inverted index, but may also have to search through the event data in the bucket associated with the bucket-specific inverted index to produce additional results on the fly. In other words, an indexer would need to search through event data stored in the bucket (that was not yet processed by the indexer for the corresponding inverted index) to supplement the partial results from the bucket-specific inverted index.
At block 742, a query is received by a data intake and query system. In some embodiments, the query can be receive as a user generated query entered into search bar of a graphical user search interface. The search interface also includes a time range control element that enables specification of a time range for the query.
At block 744, an inverted index is retrieved. Note, that the inverted index can be retrieved in response to an explicit user search command inputted as part of the user generated query. Alternatively, the search engine can be configured to automatically use an inverted index if it determines that using the inverted index would expedite the servicing of the user generated query. Each of the entries in an inverted index keeps track of instances of a specific value in a specific field in the event data and includes references to events containing the specific value in the specific field. In order to expedite queries, in most embodiments, the search engine will employ the inverted index separate from the raw record data store to generate responses to the received queries.
At block 746, the query engine determines if the query contains further filtering and processing steps. If the query contains no further commands, then, in one embodiment, summarization information can be provided to the user at block 754.
If, however, the query does contain further filtering and processing commands, then at block 750, the query engine determines if the commands relate to further filtering or processing of the data extracted as part of the inverted index or whether the commands are directed to using the inverted index as an initial filtering step to further filter and process event data referenced by the entries in the inverted index. If the query can be completed using data already in the generated inverted index, then the further filtering or processing steps, e.g., a “count” number of records function, “average” number of records per hour etc. are performed and the results are provided to the user at block 752.
If, however, the query references fields that are not extracted in the inverted index, then the indexers will access event data pointed to by the reference values in the inverted index to retrieve any further information required at block 756. Subsequently, any further filtering or processing steps are performed on the fields extracted directly from the event data and the results are provided to the user at step 758.
2.13.4. Accelerating Report Generation
In some embodiments, a data server system such as the data intake and query system can accelerate the process of periodically generating updated reports based on query results. To accelerate this process, a summarization engine automatically examines the query to determine whether generation of updated reports can be accelerated by creating intermediate summaries. If reports can be accelerated, the summarization engine periodically generates a summary covering data obtained during a latest non-overlapping time period. For example, where the query seeks events meeting a specified criteria, a summary for the time period includes only events within the time period that meet the specified criteria. Similarly, if the query seeks statistics calculated from the events, such as the number of events that match the specified criteria, then the summary for the time period includes the number of events in the period that match the specified criteria.
In addition to the creation of the summaries, the summarization engine schedules the periodic updating of the report associated with the query. During each scheduled report update, the query engine determines whether intermediate summaries have been generated covering portions of the time period covered by the report update. If so, then the report is generated based on the information contained in the summaries. Also, if additional event data has been received and has not yet been summarized, and is required to generate the complete report, the query can be run on these additional events. Then, the results returned by this query on the additional events, along with the partial results obtained from the intermediate summaries, can be combined to generate the updated report. This process is repeated each time the report is updated. Alternatively, if the system stores events in buckets covering specific time ranges, then the summaries can be generated on a bucket-by-bucket basis. Note that producing intermediate summaries can save the work involved in re-running the query for previous time periods, so advantageously only the newer events needs to be processed while generating an updated report. These report acceleration techniques are described in more detail in U.S. Pat. No. 8,589,403, entitled “Compressed Journaling In Event Tracking Files For Metadata Recovery And Replication”, issued on 19 Nov. 2013, U.S. Pat. No. 8,412,696, entitled “Real Time Searching And Reporting”, issued on 2 Apr. 2011, and U.S. Pat. Nos. 8,589,375 and 8,589,432, both also entitled “REAL TIME SEARCHING AND REPORTING”, both issued on 19 Nov. 2013, each of which is hereby incorporated by reference in its entirety for all purposes.
2.14. Security Features
The data intake and query system provides various schemas, dashboards, and visualizations that simplify developers' tasks to create applications with additional capabilities. One such application is the an enterprise security application, such as SPLUNK® ENTERPRISE SECURITY, which performs monitoring and alerting operations and includes analytics to facilitate identifying both known and unknown security threats based on large volumes of data stored by the data intake and query system. The enterprise security application provides the security practitioner with visibility into security-relevant threats found in the enterprise infrastructure by capturing, monitoring, and reporting on data from enterprise security devices, systems, and applications. Through the use of the data intake and query system searching and reporting capabilities, the enterprise security application provides a top-down and bottom-up view of an organization's security posture.
The enterprise security application leverages the data intake and query system search-time normalization techniques, saved searches, and correlation searches to provide visibility into security-relevant threats and activity and generate notable events for tracking. The enterprise security application enables the security practitioner to investigate and explore the data to find new or unknown threats that do not follow signature-based patterns.
Conventional Security Information and Event Management (SIEM) systems lack the infrastructure to effectively store and analyze large volumes of security-related data. Traditional SIEM systems typically use fixed schemas to extract data from pre-defined security-related fields at data ingestion time and store the extracted data in a relational database. This traditional data extraction process (and associated reduction in data size) that occurs at data ingestion time inevitably hampers future incident investigations that may need original data to determine the root cause of a security issue, or to detect the onset of an impending security threat.
In contrast, the enterprise security application system stores large volumes of minimally-processed security-related data at ingestion time for later retrieval and analysis at search time when a live security threat is being investigated. To facilitate this data retrieval process, the enterprise security application provides pre-specified schemas for extracting relevant values from the different types of security-related events and enables a user to define such schemas.
The enterprise security application can process many types of security-related information. In general, this security-related information can include any information that can be used to identify security threats. For example, the security-related information can include network-related information, such as IP addresses, domain names, asset identifiers, network traffic volume, uniform resource locator strings, and source addresses. The process of detecting security threats for network-related information is further described in U.S. Pat. No. 8,826,434, entitled “SECURITY THREAT DETECTION BASED ON INDICATIONS IN BIG DATA OF ACCESS TO NEWLY REGISTERED DOMAINS”, issued on 2 Sep. 2014, U.S. Pat. No. 9,215,240, entitled “INVESTIGATIVE AND DYNAMIC DETECTION OF POTENTIAL SECURITY-THREAT INDICATORS FROM EVENTS IN BIG DATA”, issued on 15 Dec. 2015, U.S. Pat. No. 9,173,801, entitled “GRAPHIC DISPLAY OF SECURITY THREATS BASED ON INDICATIONS OF ACCESS TO NEWLY REGISTERED DOMAINS”, issued on 3 Nov. 2015, U.S. Pat. No. 9,248,068, entitled “SECURITY THREAT DETECTION OF NEWLY REGISTERED DOMAINS”, issued on 2 Feb. 2016, U.S. Pat. No. 9,426,172, entitled “SECURITY THREAT DETECTION USING DOMAIN NAME ACCESSES”, issued on 23 Aug. 2016, and U.S. Pat. No. 9,432,396, entitled “SECURITY THREAT DETECTION USING DOMAIN NAME REGISTRATIONS”, issued on 30 Aug. 2016, each of which is hereby incorporated by reference in its entirety for all purposes. Security-related information can also include malware infection data and system configuration information, as well as access control information, such as login/logout information and access failure notifications. The security-related information can originate from various sources within a data center, such as hosts, virtual machines, storage devices and sensors. The security-related information can also originate from various sources in a network, such as routers, switches, email servers, proxy servers, gateways, firewalls and intrusion-detection systems.
During operation, the enterprise security application facilitates detecting “notable events” that are likely to indicate a security threat. A notable event represents one or more anomalous incidents, the occurrence of which can be identified based on one or more events (e.g., time stamped portions of raw machine data) fulfilling pre-specified and/or dynamically-determined (e.g., based on machine-learning) criteria defined for that notable event. Examples of notable events include the repeated occurrence of an abnormal spike in network usage over a period of time, a single occurrence of unauthorized access to system, a host communicating with a server on a known threat list, and the like. These notable events can be detected in a number of ways, such as: (1) a user can notice a correlation in events and can manually identify that a corresponding group of one or more events amounts to a notable event; or (2) a user can define a “correlation search” specifying criteria for a notable event, and every time one or more events satisfy the criteria, the application can indicate that the one or more events correspond to a notable event; and the like. A user can alternatively select a pre-defined correlation search provided by the application. Note that correlation searches can be run continuously or at regular intervals (e.g., every hour) to search for notable events. Upon detection, notable events can be stored in a dedicated “notable events index,” which can be subsequently accessed to generate various visualizations containing security-related information. Also, alerts can be generated to notify system operators when important notable events are discovered.
The enterprise security application provides various visualizations to aid in discovering security threats, such as a “key indicators view” that enables a user to view security metrics, such as counts of different types of notable events. For example,
These visualizations can also include an “incident review dashboard” that enables a user to view and act on “notable events.” These notable events can include: (1) a single event of high importance, such as any activity from a known web attacker; or (2) multiple events that collectively warrant review, such as a large number of authentication failures on a host followed by a successful authentication. For example,
2.15. Data Center Monitoring
As mentioned above, the data intake and query platform provides various features that simplify the developers' task to create various applications. One such application is a virtual machine monitoring application, such as SPLUNK® APP FOR VMWARE® that provides operational visibility into granular performance metrics, logs, tasks and events, and topology from hosts, virtual machines and virtual centers. It empowers administrators with an accurate real-time picture of the health of the environment, proactively identifying performance and capacity bottlenecks.
Conventional data-center-monitoring systems lack the infrastructure to effectively store and analyze large volumes of machine-generated data, such as performance information and log data obtained from the data center. In conventional data-center-monitoring systems, machine-generated data is typically pre-processed prior to being stored, for example, by extracting pre-specified data items and storing them in a database to facilitate subsequent retrieval and analysis at search time. However, the rest of the data is not saved and discarded during pre-processing.
In contrast, the virtual machine monitoring application stores large volumes of minimally processed machine data, such as performance information and log data, at ingestion time for later retrieval and analysis at search time when a live performance issue is being investigated. In addition to data obtained from various log files, this performance-related information can include values for performance metrics obtained through an application programming interface (API) provided as part of the vSphere Hypervisor™ system distributed by VMware, Inc. of Palo Alto, Calif. For example, these performance metrics can include: (1) CPU-related performance metrics; (2) disk-related performance metrics; (3) memory-related performance metrics; (4) network-related performance metrics; (5) energy-usage statistics; (6) data-traffic-related performance metrics; (7) overall system availability performance metrics; (8) cluster-related performance metrics; and (9) virtual machine performance statistics. Such performance metrics are described in U.S. patent application Ser. No. 14/167,316, entitled “Correlation For User-Selected Time Ranges Of Values For Performance Metrics Of Components In An Information-Technology Environment With Log Data From That Information-Technology Environment”, filed on 29 Jan. 2014, and which is hereby incorporated by reference in its entirety for all purposes.
To facilitate retrieving information of interest from performance data and log files, the virtual machine monitoring application provides pre-specified schemas for extracting relevant values from different types of performance-related events, and also enables a user to define such schemas.
The virtual machine monitoring application additionally provides various visualizations to facilitate detecting and diagnosing the root cause of performance problems. For example, one such visualization is a “proactive monitoring tree” that enables a user to easily view and understand relationships among various factors that affect the performance of a hierarchically structured computing system. This proactive monitoring tree enables a user to easily navigate the hierarchy by selectively expanding nodes representing various entities (e.g., virtual centers or computing clusters) to view performance information for lower-level nodes associated with lower-level entities (e.g., virtual machines or host systems). Example node-expansion operations are illustrated in
The virtual machine monitoring application also provides a user interface that enables a user to select a specific time range and then view heterogeneous data comprising events, log data, and associated performance metrics for the selected time range. For example, the screen illustrated in
2.16. IT Service Monitoring
As previously mentioned, the data intake and query platform provides various schemas, dashboards and visualizations that make it easy for developers to create applications to provide additional capabilities. One such application is an IT monitoring application, such as SPLUNK® IT SERVICE INTELLIGENCE™, which performs monitoring and alerting operations. The IT monitoring application also includes analytics to help an analyst diagnose the root cause of performance problems based on large volumes of data stored by the data intake and query system as correlated to the various services an IT organization provides (a service-centric view). This differs significantly from conventional IT monitoring systems that lack the infrastructure to effectively store and analyze large volumes of service-related events. Traditional service monitoring systems typically use fixed schemas to extract data from pre-defined fields at data ingestion time, wherein the extracted data is typically stored in a relational database. This data extraction process and associated reduction in data content that occurs at data ingestion time inevitably hampers future investigations, when all of the original data may be needed to determine the root cause of or contributing factors to a service issue.
In contrast, an IT monitoring application system stores large volumes of minimally-processed service-related data at ingestion time for later retrieval and analysis at search time, to perform regular monitoring, or to investigate a service issue. To facilitate this data retrieval process, the IT monitoring application enables a user to define an IT operations infrastructure from the perspective of the services it provides. In this service-centric approach, a service such as corporate e-mail may be defined in terms of the entities employed to provide the service, such as host machines and network devices. Each entity is defined to include information for identifying all of the events that pertains to the entity, whether produced by the entity itself or by another machine, and considering the many various ways the entity may be identified in machine data (such as by a URL, an IP address, or machine name). The service and entity definitions can organize events around a service so that all of the events pertaining to that service can be easily identified. This capability provides a foundation for the implementation of Key Performance Indicators.
One or more Key Performance Indicators (KPI's) are defined for a service within the IT monitoring application. Each KPI measures an aspect of service performance at a point in time or over a period of time (aspect KPI's). Each KPI is defined by a search query that derives a KPI value from the machine data of events associated with the entities that provide the service. Information in the entity definitions may be used to identify the appropriate events at the time a KPI is defined or whenever a KPI value is being determined. The KPI values derived over time may be stored to build a valuable repository of current and historical performance information for the service, and the repository, itself, may be subject to search query processing. Aggregate KPIs may be defined to provide a measure of service performance calculated from a set of service aspect KPI values; this aggregate may even be taken across defined timeframes and/or across multiple services. A particular service may have an aggregate KPI derived from substantially all of the aspect KPI's of the service to indicate an overall health score for the service.
The IT monitoring application facilitates the production of meaningful aggregate KPI's through a system of KPI thresholds and state values. Different KPI definitions may produce values in different ranges, and so the same value may mean something very different from one KPI definition to another. To address this, the IT monitoring application implements a translation of individual KPI values to a common domain of “state” values. For example, a KPI range of values may be 1-100, or 50-275, while values in the state domain may be ‘critical,’ ‘warning,’ ‘normal,’ and ‘informational’. Thresholds associated with a particular KPI definition determine ranges of values for that KPI that correspond to the various state values. In one case, KPI values 95-100 may be set to correspond to ‘critical’ in the state domain. KPI values from disparate KPI's can be processed uniformly once they are translated into the common state values using the thresholds. For example, “normal 80% of the time” can be applied across various KPI's. To provide meaningful aggregate KPI's, a weighting value can be assigned to each KPI so that its influence on the calculated aggregate KPI value is increased or decreased relative to the other KPI's.
One service in an IT environment often impacts, or is impacted by, another service. The IT monitoring application can reflect these dependencies. For example, a dependency relationship between a corporate e-mail service and a centralized authentication service can be reflected by recording an association between their respective service definitions. The recorded associations establish a service dependency topology that informs the data or selection options presented in a GUI, for example. (The service dependency topology is like a “map” showing how services are connected based on their dependencies.) The service topology may itself be depicted in a GUI and may be interactive to allow navigation among related services.
Entity definitions in the IT monitoring application can include informational fields that can serve as metadata, implied data fields, or attributed data fields for the events identified by other aspects of the entity definition. Entity definitions in the IT monitoring application can also be created and updated by an import of tabular data (as represented in a CSV, another delimited file, or a search query result set). The import may be GUI-mediated or processed using import parameters from a GUI-based import definition process. Entity definitions in the IT monitoring application can also be associated with a service by means of a service definition rule. Processing the rule results in the matching entity definitions being associated with the service definition. The rule can be processed at creation time, and thereafter on a scheduled or on-demand basis. This allows dynamic, rule-based updates to the service definition.
During operation, the IT monitoring application can recognize notable events that may indicate a service performance problem or other situation of interest. These notable events can be recognized by a “correlation search” specifying trigger criteria for a notable event: every time KPI values satisfy the criteria, the application indicates a notable event. A severity level for the notable event may also be specified. Furthermore, when trigger criteria are satisfied, the correlation search may additionally or alternatively cause a service ticket to be created in an IT service management (ITSM) system, such as a systems available from ServiceNow, Inc., of Santa Clara, Calif.
SPLUNK® IT SERVICE INTELLIGENCE™ provides various visualizations built on its service-centric organization of events and the KPI values generated and collected.
Visualizations can be particularly useful for monitoring or investigating service performance. The IT monitoring application provides a service monitoring interface suitable as the home page for ongoing IT service monitoring. The interface is appropriate for settings such as desktop use or for a wall-mounted display in a network operations center (NOC). The interface may prominently display a services health section with tiles for the aggregate KPI's indicating overall health for defined services and a general KPI section with tiles for KPI's related to individual service aspects. These tiles may display KPI information in a variety of ways, such as by being colored and ordered according to factors like the KPI state value. They also can be interactive and navigate to visualizations of more detailed KPI information.
The IT monitoring application provides a service-monitoring dashboard visualization based on a user-defined template. The template can include user-selectable widgets of varying types and styles to display KPI information. The content and the appearance of widgets can respond dynamically to changing KPI information. The KPI widgets can appear in conjunction with a background image, user drawing objects, or other visual elements, that depict the IT operations environment, for example. The KPI widgets or other GUI elements can be interactive so as to provide navigation to visualizations of more detailed KPI information.
The IT monitoring application provides a visualization showing detailed time-series information for multiple KPI's in parallel graph lanes. The length of each lane can correspond to a uniform time range, while the width of each lane may be automatically adjusted to fit the displayed KPI data. Data within each lane may be displayed in a user selectable style, such as a line, area, or bar chart. During operation a user may select a position in the time range of the graph lanes to activate lane inspection at that point in time. Lane inspection may display an indicator for the selected time across the graph lanes and display the KPI value associated with that point in time for each of the graph lanes. The visualization may also provide navigation to an interface for defining a correlation search, using information from the visualization to pre-populate the definition.
The IT monitoring application provides a visualization for incident review showing detailed information for notable events. The incident review visualization may also show summary information for the notable events over a timeframe, such as an indication of the number of notable events at each of a number of severity levels. The severity level display may be presented as a rainbow chart with the warmest color associated with the highest severity classification. The incident review visualization may also show summary information for the notable events over a timeframe, such as the number of notable events occurring within segments of the timeframe. The incident review visualization may display a list of notable events within the timeframe ordered by any number of factors, such as time or severity. The selection of a particular notable event from the list may display detailed information about that notable event, including an identification of the correlation search that generated the notable event.
The IT monitoring application provides pre-specified schemas for extracting relevant values from the different types of service-related events. It also enables a user to define such schemas.
2.17 Cloud-Based Architecture
As shown in the previous figures, various embodiments may refer to a data intake and query system 108 that includes one or more of a search head 210, an indexer 206, and a forwarder 204. In other implementations, data intake and query system 108 may have a different architecture, but may carry out indexing and searching in a way that is indistinguishable or functionally equivalent from the perspective of the end user. For example, data intake and query system 108 may be re-architected to run in a stateless, containerized environment. In some of these embodiments, data intake and query system 108 may be run in a computing cloud provided by a third party, or provided by the operator of the data intake and query system 108. This type of cloud-based data intake and query system may have several benefits, including, but not limited to, lossless data ingestion, more robust disaster recovery, and faster or more efficient processing, searching, and indexing. A cloud-based data intake and query system as described in this section may provide separately scalable storage resources and compute resources, or separately scalable search and index resources. Additionally, the cloud-based data intake and query system may allow for applications to be developed on top of the data intake and query system, to extend or enhance functionality, through a gateway layer or one or more Application Programming Interfaces (APIs), which may provide customizable access control or targeted exposure to the workings of data intake and query system 108.
In some embodiments, a cloud-based data intake and query system may include an intake system. Such an intake system can include, but is not limited to an intake buffer, such as Apache Kafka (R) or Amazon Kinesis (R), or an extensible compute layer, such as Apache Spark™ or Apache Flink (R). In some embodiments, the search function and the index function may be separated or containerized, so that search functions and index functions may run or scale independently. In some embodiments, data that is indexed may be stored in buckets, which may be stored in a persistent storage once certain bucket requirements have been met, and retrieved as needed for searching. In some embodiments, the search functions and index functions run in stateless containers, which may be coordinated by an orchestration platform. These containerized search and index functions may retrieve data needed to carry out searching and indexing from the buckets or various other services that may also run in containers, or within other components of the orchestration platform. In this manner, loss of a single container, or even multiple containers, does not result in data loss, because the data can be quickly recovered from the various services or components or the buckets in which the data is persisted.
In some embodiments, the cloud-based data intake and query system may implement tenant-based and user-based access control. In some embodiments, the cloud-based data intake and query system may implement an abstraction layer, through a gateway portal, an API, or some combination thereof, to control or limit access to the functionality of the cloud-based data intake and query system.
2.18 HTTP-Based Data Submission
A common trend in the development of computer-implemented technology is standardization, whereby different systems are enabled to interoperate by a standard protocol, interface, or the like. For example, the majority of the current world-wide-web (WWW) operates according to a version of the hypertext transport protocol (HTTP), a protocol known in the art. Given the prevalence of HTTP, it may be desirable for the data intake and query system described above to enable submission of data via HTTP.
One example embodiment of a data intake and query system 1808 that enables submission of data via HTTP is shown in
Unlike the data intake and query system 108 of
One example of an HTTP-request from a data source 202 to an HEC 1806 is a request to submit data to the data intake and query system 1808. Illustratively, the HEC 1806 may implement an application programming interface (API) to which such requests can be submitted. In one embodiment, the API conforms to the representational state transfer (REST) architectural style, which is known in the art. Pursuant to the REST architectural style, data sources 202 may therefore interact with the HECs 1806 via a predefined set of stateless operations. Illustratively, a POST operation may be used to submit data to the HEC 1806 for processing. The POST operation may be directed to a uniform resource identifier (URI) associated with the HEC 1806 (e.g., including a domain name associated with the system 1808 and a transport control protocol (TCP) port on which the HEC 1806 implements the API). The POST operation may include a payload containing the data from the data source 202 to be processed by the system 1808 (e.g., to be indexed by an indexer 206 and persisted in the data store 208). In one embodiment, the payload includes unstructured, unprocessed, or otherwise raw machine data. In another embodiment, the payload includes structured data. The structured data may include, for example, key-value pairs represented according to the JavaScript Object Notation (JSON) standard, which standard is known in the art.
On receiving a request to process data, an HEC 1806 can submit the data to forwarder 204 to be processed, in accordance with the embodiments described above. In some instances, the HEC 1806 may be integrated with the forwarder 204, for example by operating as a software application on a host device implementing the forwarder 204. In some embodiments, the data intake and query system 1808 may operate without forwarders 204, and thus the HECs 1806 may submit data to an indexer 206, rather than a forwarder 204. In some instances, the HEC 1806 may be integrated with the indexer 206. As discussed above, on receiving data from a data source 202, the indexer 208 can proceed to process the data, such as by indexing the data and storing it within the data store 208. In one embodiment, the indexer 206 maintains not-yet processed data within a non-persistent data cache (e.g., implemented in random access memory (RAM)). Thus, processing data may correspond to storing of the data within the data store 208, illustratively representing persistent memory (e.g., a hard disk drive) of the system 1808.
In one embodiment, the HECs 1806 interact with data sources 202 via HTTP or another standardized communication protocol, and with the indexers 206 and/or forwarders 204 via a proprietary protocol (e.g., proprietary to the system 1808). Thus, HECs 1806 may be conceptually viewed as an interface between a standardized protocol (e.g., HTTP) and the proprietary protocol.
2.18.1 Stateless Data Ingestion
As the number of network devices that collect data grows at an exponential rate, modifications may be made to the HECs 1806 to account for the mass amount of data being ingested via HTTP. As referenced above, some embodiments of the data intake and query system 108 include one or more indexers 206 that maintain not-yet processed data within a non-persistent data cache (e.g., implemented in random access memory (RAM) or other volatile memory). However, due to the mass amounts of data ingested via HTTP by some implementations of the data intake and query system 108, a persistent queue may be included within a data intake and query system 108 to improve the data intake process via HTTP. The inclusion of the persistent queue, as will be discussed below, address issues with the current implementation of the data intake and query system 108 involving data loss, a sticky acknowledgement poll, data duplication, and a non-idempotent acknowledgement poll.
Additionally, it is often advantageous for a data source to receive an acknowledgement that data transmitted to a data intake and query system has been received and stored. However, in previous technologies, ingested data may be stored in volatile memory and be assigned an identifier (“dataID”). Further, the data source may have been provided an initial acknowledgment indicating receipt of the data along with the dataID. The data source was then tasked with querying the data intake and query system for the status of the data using the dataID (namely whether the data had been successfully stored). The process of providing an initial acknowledgement with a dataID and requiring the data source to query the data intake and query system for the status of the data lead to several issues, which included data loss, data duplication, a sticky acknowledgment poll problem, and a non-idempotent acknowledgement poll problem.
Data loss occurred as a result of software and/or hardware failure associated with the data intake and query system, or a loss of power. For instance, as ingested data is often stored in a cache (e.g., RAM or other volatile memory), loss of power or a data intake and query system crash would result in the loss of the data in the cache. In situations where the data is not able to be replicated by the data source, data loss occurred.
The sticky acknowledgement poll problem occurred due to the fact that a plurality of instances of a data intake and query system may operate within a single network (e.g., a multiple instance configuration within an enterprise network). In previous technologies, as a single data intake and query system instance maintains a list of dataID's corresponding to ingested data being parsed, indexed and stored (or in queue to be parsed, indexed and stored), only the particular data intake and query system that ingested the data has knowledge of the status of the data. As a result, a data source had to query the particular data intake and query system instance that ingested the data (e.g., the “sticky acknowledgement poll problem”).
The non-idempotent acknowledgement poll problem occurred as a result of the need by previous technologies to store the dataID at the data intake and query system even after the corresponding ingested data has been indexed and stored in order to enable the data source to query for the status of the data (i.e., to ensure the data was successfully parsed and stored). However, at times following successful parsing and storage of the data, the dataID would be deleted prior to querying by the data source due to a loss of power to the data intake and query system or the system crashed (as the list of dataID's were maintained in volatile memory). At other times, a dataID was deleted due to constraints on the number of dataID's that could be stored. Thus, even though the ingested data was successfully parsed and stored, a query for the status of the data returned a negative confirmation as the dataID was not found by the data intake and query system. Following the improper negative confirmation (e.g., the non-idempotent acknowledgement poll problem), the data source may retransmit the data for ingestion (if available), resulting in data duplication.
Referring now to
The system 1900 includes a plurality of data intake and query systems 1902 communicatively coupled to data sources 202 and a network 104. As shown in
Upon receiving data from a data source 202, instead of forwarding the ingested data to an indexer 206 as discussed in the above embodiments, a forwarder/HEC 1904 forwards the ingested data to a persistent queue 1906. The persistent queue 1906 stores the data on a corresponding disk 1908. Upon completion of storing the data on the corresponding disk 1908, the persistent queue 1906 provides an indication of such to the forwarder/HEC 1904, which transmits an acknowledgement of storage to the data source 202 that provided the data.
Providing the ingested data to the persistent queue 1906, writing the ingested data to the disk 1908 and providing an acknowledgment following successful storage to the disk 1908 solves several of the problems of previous technologies discussed above. Providing an acknowledgement that data has been successfully stored on the disk 1908 removes the need for a data source 202 to query the data intake and query system 1902 for the status of the data. As a result, there is no possibility for an acknowledgement providing an improper negative confirmation, which solves the non-idempotent acknowledgement poll and data duplication problems. Additionally, as the data source 202 no longer queries the data intake and query system 1902 for the status of ingested data, the sticky acknowledgement poll problem has been solved. Further, transmission of an acknowledgement of storage occurs following storage of the ingested data on non-volatile memory. Therefore, if the data intake and query system 1902 crashes or a loss of power occurs following storage, on the disk 1908, the data intake and query system 1902 may be able to read the data from the disk 1908 upon reinitiating, preventing data loss.
Referring to
In some embodiments, a load balancer (not shown) may be included within the data intake and query system 1902 and be located along a communication path between the data sources 202 and the forwarder/HEC 1904. The load balancer may ingest data and use traffic load balancing to route the ingested data to a particular forwarder/HEC 1904. Although any suitable load balancer may be implemented in any suitable manner, in an exemplary embodiment a load balancer such NGINX may distribute messages between event collectors at regular intervals.
Referring to
With reference to the illustration, the forwarder/HEC 1904 of the data intake and query system 1902 receives data (numeral 1). Receipt of the data may correspond to ingesting data through receipt of a POST request message from the data source 202 with the POST request message including a payload containing the data to be processed by the system 1902 (e.g., to be indexed by an indexer 206 and persisted in the data store 208). In other embodiments a load balancer may ingest the data and provide the ingested data to the forwarder/HEC 1904.
Following receipt of the ingested data, the forwarder/HEC 1904 provides the ingested data to a persistent queue 1906 (numeral 2). The persistent queue 1906 writes the ingested data, or a copy thereof, to the disk 1908 (numeral 3). The use of the term “ingested data” throughout the disclosure may refer to either the actual ingested data or a copy thereof. Following the completion of writing the ingested data to the disk 1908, the forwarder/HEC 1904 is notified by the persistent queue 1906 and transmits an acknowledgement message to the data source 202 (numeral 4). Receipt of the acknowledge message by the data source 202 confirms that the ingested data has been stored.
Once the persistent queue 1906 has written the ingested data to the disk 1908, the ingested data, is provided to the data pipeline queues 2102 (numeral 5). The data pipeline queues 2102 may include several components including, but not limited or restricted to, a parsing queue, a merging queue (or aggregation queue), and/or a typing queue. For example, the parsing queue may perform encoding (e.g., Unicode Transformation Format-8 bit, “UTF-8” encoding), linebreaking, and header processing. The merging queue may perform line merging and timestamp extraction. The typing queue may perform regex (rational expression) replacement and punctuation extraction. The data pipeline queues 2102 may perform some or all of these operations in order to prepare the ingested data for indexing and storage within the data store 208 as time-based events.
Upon completing the processing performed by the data pipeline queues 2102, the indexer 206 receives the processed data (numeral 6). The indexer 206 parses the data to organize the data into events, which may include determining a source type (e.g., by extracting a source type label from the metadata fields associated with the data, etc.) and referring to a source type configuration corresponding to the identified source type. Additionally, the indexer 206 may determine a timestamp for each event and associate each event one or more metadata fields including a field containing the timestamp determined for the event. The indexer then stores the indexed data in the data store 208 (numeral 7).
Following completion of the indexing and storage of the ingested data, the ingested data is deleted from the disk 1908 (numeral 8). As is discussed below, upon writing the indexed data to the data store 208, the indexer 206 provides an indication that the data has been indexed and stored, which initiates deletion of the ingested data from the disk 1908.
Referring to
The socket server 2202 is configured to listen to a network port of the forwarder/HEC 1904 and read ingested data from a data handler of the forwarder/HEC 1904. The ingested data is then provided to one of the tubes 22041-22042. In general, a “tube” is a data store that receives the ingested data and organizes the ingested data to be written to (or access for storage by) the disk 2206. In some embodiments, the ingested data is provided to one of the tubes 22041-22042 in “blocks,” wherein each block is a specified size. For example, the ingested data may be broken into blocks of 64 kilobytes (kB or KB). However, the size of a block is not limited to 64 KB, but rather, the size of a block may have any specified size. Additionally, each block is annotated with metadata keys that may include an indication of the host, the source, and the source type of the data. The metadata keys may also include values that indicate the character encoding of the data stream from which the data block was ingested, and values that direct the data block to a specific indexer 206.
According to one embodiment of the disclosure, each of the one or more tubes 22041-22042 may be configured as a first-in, first-out (FIFO) queue that receives blocks of ingested data from the forwarder/HEC 1904. When a block of data is first received by a tube 22041, a status of the block of data is marked as “ready” and the block of data is written to the disk 2206 by one or more components or logic of the data intake and query system 1902. In some embodiments, the status of the block may be included in the metadata. Once the process thread 2208 has reserved the block of data to be provided to the data pipeline queues 2102, the status of the block of the data is modified to be “reserved.” The reserved block of data may be provided to the data pipeline queues (or a delay may occur following the reservation in order to provide several blocks of data to the data pipeline queues at once).
After the indexer 206 has indexed the block of data, the indexer 206 modifies the status of the data to be “processed.” In some embodiments, when the status of a block of data is marked as “reserved,” and is not processed for a predetermined amount of time (e.g., the status is not modified to “processed”), the status of the block of data will revert back to “ready” so that the process thread 2208 may reserve the block of data again and provide the block of data to the data pipeline queues 2102 for processing and eventual indexing by the indexer 206.
As referenced above, the process thread 2208 periodically (or at aperiodic intervals) queries the one or more tubes 22041-22042 and reads data from a tube 2204 (e.g., tube 22041) by “reserving” a block of data marked as “ready” within the tube 22041. The block of data is then provided by the process thread 2208 to the data pipeline queues 2102. The block of data is processed by the components of the data pipeline queues 2102 (e.g., the parsing queue, the merging queue, and the typing queue) as discussed above. Following processing by the data pipeline queues 2102, the processed block of data is indexed by the indexer 206 and stored in the data store 208. Upon completion of indexing and storage of the block of data, the indexer 206 modifies the status of the block of data to be “processed,” and the block of data is deleted from the disk 2206.
Referring to
Following receipt of the ingested data, the forwarder/HEC 1904 provides the ingested data to a first persistent queue 1906 (numeral 2A) and an additional forwarder/HEC instance 2302 (numeral 2C). Providing the ingested data to the first persistent queue 1906 represents the first level of persistence while providing the ingested data to the second persistent queue 2302 represents the second level of persistence, wherein the second level of persistence refers to a second, duplicate copy of the ingested data being stored on persistent storage until completion of the indexing and storage process.
The persistent queue 1906 writes the ingested data to the disk 1908 (numeral 2B) while the additional forwarder/HEC instance 2302 provides the data to a secondary persistent queue (numeral 2C), which writes the ingested data to the secondary disk 2306 (numeral 2D). It should be noted that the reference numerals 2A-2E do not refer to a strict ordering of operations. Instead, the operation of numeral 2A may occur before the operation of numeral 2B, and the operation of numeral 2C may occur before the operation of numeral 2D, which occurs before the operation of numeral 2E; however, the operations of numerals 2A-2B may occur before, after, concurrently or simultaneously with the operations of numerals 2C-2E. Following the completion of writing the ingested data to the disk 1908 and the secondary disk 2306, the forwarder/HEC 1904 transmits an acknowledgement that the ingested data has been stored to the data source 202 (numeral 3).
Once the ingested data has been written to the disk 1908 and the secondary disk 2312, the ingested data is provided to the data pipeline queues 2102 (numeral 4). The data pipeline queues 2102 may perform processing by one or more several components including the parsing queue, the merging queue and the typing queue, as discussed above. Upon completing the processing performed by the data pipeline queues 2102, the indexer 206 receives the processed ingested data (numeral 5). The indexer 206 parses the data to organize the data into events and stores the indexed data in the data store 208 (numeral 6).
Following completion of the indexing and storage of the ingested data, the ingested data is deleted from the disk 1908 (numeral 7A) and from the secondary disk 2306 (numeral 7B). As is discussed below, upon writing the indexed data to the data store 208, the indexer 206 provides an indication that the data has been indexed and stored, which initiates the deleting of the ingested data from the disk 1908 and from the secondary disk 2306.
Referring to
Following receipt of the ingested data, the forwarder/HEC 1904 provides the ingested data to a first persistent queue 1906 (numeral 2A) and an additional forwarder/HEC instance 2302 (numeral 2E). Providing the ingested data to the first persistent queue 1906 represents the first level of persistence while providing the ingested data to the second persistent queue 2302 represents the second level of persistence, wherein the second level of persistence refers to a second, duplicate copy of the ingested data being stored on persistent storage until completion of the indexing and storage process.
The socket server 2202 of the persistent queue 1906 reads the ingested data from the forwarder/HEC 1906 (numeral 2A). The socket server 2202 provides a tube 2204i with one or more blocks of ingested data from the forwarder/HEC 1904 (numeral 2B). In some embodiments, multiple tubes 22041-2204i may each be provided a portion of the ingested data. The status of each of the one or more blocks of data is set to “ready,” and thereafter, the block is written to the disk 2206 (numeral 2C). An indication that the ingested data has been stored on the disk 2206 is provided to the forwarder/HEC 1904 (numeral 2D). The “indication” may include a message, such as an acknowledgement message, albeit other forms of indication (e.g., signal, setting tags that are read by logic within the forwarder/HEC 1904, or any other mechanisms to convey storage of the ingested data on disk 2206) may be used.
Additionally, the ingested data is provided by the forwarder/HEC 1904 to the additional forwarder/HEC instance 2302 (numeral 2E). The socket server 2402 of the secondary persistent queue 2304 reads the ingested data from the forwarder/HEC 2302 (numeral 2F). The socket server 2402 provides the tube 24041 with one or more blocks of ingested data (numeral 2G). In some embodiments, multiple tubes 24041-2404j (wherein j≥3) may each be provided a portion of the ingested data. The status of each of the one or more blocks of data is set to “ready” and thereafter, the block is written to the secondary disk 2406 (numeral 2H) and provided to the tube 24043, which stores an identifier assigned to the ingested data along with the ingested data in a raw form (e.g., as received from the data source 202). An indication that the ingested data has been stored on the secondary disk 2406 is provided to the forwarder/HEC 1904 (numeral 2I). In a similar manner as discussed above with respect to the ordering of operations of numerals 2A-2E of
Following the completion of writing the ingested data to the disk 2206 and the secondary disk 2406, the forwarder/HEC 1904 transmits an acknowledgement to the data source 202 that the ingested data has been stored (numeral 3).
Once the ingested data has been written to the disk 2206 and the secondary disk 2406, the ingested data is provided to the data pipeline queues 2102 by the process thread 2208 and/or 2408 (numeral 4). The data pipeline queues 2102 may perform processing by one or more several components including the parsing queue, the merging queue and the typing queue, as discussed above. Upon completing the processing performed by the data pipeline queues 2102, the indexer 206 receives the processed ingested data (numeral 5). The indexer 206 parses the data to organize the data into events and stores the indexed data in the data store 208 (numeral 6).
Following completion of the indexing and storage of the ingested data, the indexer 206 provides the identifier assigned to the ingested data to the tube 22042. The recovery thread 2410 routinely reads the contents of the tube 22042 (e.g., at regular intervals). When the identifier assigned to the ingested data is received indicating completion of the indexing and storing of the ingested data, the ingested data having the assigned identifier in a second tube is marked as completed in the tube 22043. The ingested data is deleted from the tube 22043, from the disk 2206 (numeral 7A) and from the secondary disk 2406 (numeral 7B). As is discussed below, upon writing the indexed data to the data store 208, the indexer 206 provides an indication that the data has been indexed and stored, which initiates the deleting of the ingested data from the disk 1908 and from the disk 2406.
As noted above, the recovery thread 2410 continually reads the contents of the tube 22042 checking for identifiers provided by the indexer 206. In the situation in which the first instance of the persistent queue (or any component of the data pipeline queues 2102 or the indexer 206) is unable to complete the indexing process for a particular block of ingested data, the identifier assigned to that block of ingested data will not be received by the tube 22042. Following expiration of a predetermined time after receipt of the block of ingested data, if the identifier assigned to the block of ingested data is not be received, the recovery thread 2410 moves, or causes movement of, the block of ingested data from the tube 22043 to the tube 22042 so that it may be provided to the data pipeline queues 2102 again. The predetermined time may be configurable via input from an administrator or other user. Additionally, when the data intake and query system 1902 fails (e.g., crashes or a power source is disconnected), the ingested data that was received by the persistent queue 1906 and the persistent queue 2304 but failed to be completely indexed and stored in the data store 208 is read from the disk 2206 and/or the disk 2406, as such is non-volatile storage and will maintain the ingested data stored thereon even following failure of the data intake and query system 1902. It should be noted that in some embodiments, the tubes 22041-2204j may communicate to exchange data. Similarly, in some embodiments, the tubes 24041-2404j may communicate to exchange data.
Referring to
Upon receipt of the data to be ingested, the forwarder/HEC 1904 provides the data as received to a persistent queue 1906, which writes a copy of the ingested data to the disk 1908 (block 2504). As is illustrated in
Following storage of the ingested data on disk 1908 by the persistent queue 1906, an acknowledgement message is transmitted from the forwarder/HEC 1904 to the data source 202 that submitted the data for ingestion (block 2506). In some embodiments, in lieu of previous technologies, the acknowledgement message need not include an identifier assigned to the ingested data. As discussed above, the above-described embodiments of the data intake and query system 1902 overcome the sticky acknowledgement poll problem, data duplication, and the non-idempotent acknowledgement poll problem by avoiding asynchronous confirmation. Thus, in some embodiments, the acknowledgement message provided by the forwarder/HEC 1904 confirms the storage of the data without the need for the data source 202 to later check on the status of processing of the data (as described above with respect to the non-idempotent acknowledgement poll and sticky acknowledgement problems).
Also subsequent to writing of a copy of the ingested data to the disk 1908, the data is provided to an indexer 206 by the persistent queue (block 2508). The operations performed in blocks 2506 and 2508 may be performed in any order and may be performed concurrently (e.g., at least partially overlapping in time) or simultaneously.
Upon receipt of the data at the indexer 206, the data is processed as discussed above with respect to at least
Following the processing of the ingested data by the indexer 206, the indexed data is stored in a data store 208 (block 2512). The method 2500 continues following the processing of the ingested data and storage of the indexed data in a data store 208 as the copy of the ingested data is deleted from the disk 1908 (block 2514). As referenced above with respect to
In the foregoing description, the invention is described with reference to specific exemplary embodiments thereof. It will, however, be evident that various modifications and changes may be made thereto without departing from the broader spirit and scope of the invention as set forth in the appended claims.
This application claims the benefit of priority to U.S. Provisional Application No. 62/924,052, filed Oct. 21, 2019, the entire contents of which are incorporated by reference herein.
| Number | Name | Date | Kind |
|---|---|---|---|
| 10248561 | Garvey | Apr 2019 | B2 |
| 20180027006 | Zimmermann | Jan 2018 | A1 |
| 20180091529 | Luger | Mar 2018 | A1 |
| 20190028376 | Ganapathy | Jan 2019 | A1 |
| Number | Date | Country | |
|---|---|---|---|
| 62924052 | Oct 2019 | US |
