Computing systems for querying of large sets of data can be extremely difficult to implement and maintain. In many scenarios, for example, it is necessary to first create and configure the infrastructure (e.g. server computers, storage devices, networking devices, etc.) to be used for the querying operations. It might then be necessary to perform extract, transform, and load (“ETL”) operations to obtain data from a source system and place the data in data storage. It can also be complex and time consuming to install, configure, and maintain the database management system (“DBMS”) that performs the query operations. Moreover, many DBMS are not suitable for querying extremely large data sets in a performant manner.
Computing clusters can be utilized in some scenarios to query large data sets in a performant manner. For instance, a computing cluster can have many nodes that each execute a distributed query framework for performing distributed querying of a large data set. Such computing clusters and distributed query frameworks are, however, also difficult to implement, configure, and maintain. Moreover, incorrect configuration and/or use of computing clusters such as these can result in the non-optimal utilization of processor, storage, network and, potentially, other types of computing resources.
The disclosure made herein is presented with respect to these and other considerations.
While embodiments are described herein by way of example for several embodiments and illustrative drawings, those skilled in the art will recognize that embodiments are not limited to the embodiments or drawings described. It should be understood, that the drawings and detailed description thereto are not intended to limit embodiments to the particular form disclosed, but on the contrary, the intention is to cover all modifications, equivalents and alternatives falling within the spirit and scope as defined by the appended claims. The headings used herein are for organizational purposes only and are not meant to be used to limit the scope of the description or the claims. As used throughout this application, the word “may” is used in a permissive sense (i.e., meaning having the potential to), rather than the mandatory sense (i.e., meaning must). Similarly, the words “include,” “including,” and “includes” mean including, but not limited to.
It will also be understood that, although the terms first, second, etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, a first contact could be termed a second contact, and, similarly, a second contact could be termed a first contact, without departing from the scope of the present invention. The first contact and the second contact are both contacts, but they are not the same contact.
Various embodiments of selecting resource configurations for query execution are described herein.
For example, as illustrated in
A resource may be selected, as indicated at 140, that satisfies the execution limitation (e.g., a resource selection goal), and may be instructed to perform the query 112. For example, selected computing resource 140 may implement a cluster of nodes that apply a SQL query to one or multiple data sets stored in one or multiple locations in order to generate query results 142. In addition to generating query results 142, query execution data 116 may be collected and stored as part of query execution history 130, in order to update the resource selection data available for analysis at resource selection 110 with further examples of resource configuration mapped to a performance outcome for a query, as discussed below with regard to
Please note that the previous description of selecting resource configurations for query execution is a logical illustration and thus is not to be construed as limiting as to the implementation of resource selection, computing resources, or query execution history.
This specification begins with a general description of a provider network that implements a resource management service that selecting resource configurations for query execution that are queries received from another network-based service, a managed query service. Then various examples of the managed query service and resource management service (along with other services that may be utilized or implemented) including different components/modules, or arrangements of components/module that may be employed as part of implementing the services are discussed. A number of different methods and techniques to implement selecting resource configurations for query execution are then discussed, some of which are illustrated in accompanying flowcharts. Finally, a description of an example computing system upon which the various components, modules, systems, devices, and/or nodes may be implemented is provided. Various examples are provided throughout the specification.
In various embodiments, the components illustrated in
Virtual compute service 210 may be implemented by provider network 200, in some embodiments. Virtual computing service 210 may offer instances and according to various configurations for client(s) 250 operation. A virtual compute instance may, for example, comprise one or more servers with a specified computational capacity (which may be specified by indicating the type and number of CPUs, the main memory size, and so on) and a specified software stack (e.g., a particular version of an operating system, which may in turn run on top of a hypervisor). A number of different types of computing devices may be used singly or in combination to implement the compute instances and of provider network 200 in different embodiments, including general purpose or special purpose computer servers, storage devices, network devices and the like. In some embodiments instance client(s) 250 or other any other user may be configured (and/or authorized) to direct network traffic to a compute instance.
Compute instances may operate or implement a variety of different platforms, such as application server instances, Java™ virtual machines (JVMs), general purpose or special-purpose operating systems, platforms that support various interpreted or compiled programming languages such as Ruby, Perl, Python, C, C++ and the like, or high-performance computing platforms) suitable for performing client(s) 202 applications, without for example requiring the client(s) 250 to access an instance. Applications (or other software operated/implemented by a compute instance and may be specified by client(s), such as custom and/or off-the-shelf software.
In some embodiments, compute instances have different types or configurations based on expected uptime ratios. The uptime ratio of a particular compute instance may be defined as the ratio of the amount of time the instance is activated, to the total amount of time for which the instance is reserved. Uptime ratios may also be referred to as utilizations in some implementations. If a client expects to use a compute instance for a relatively small fraction of the time for which the instance is reserved (e.g., 30%-35% of a year-long reservation), the client may decide to reserve the instance as a Low Uptime Ratio instance, and pay a discounted hourly usage fee in accordance with the associated pricing policy. If the client expects to have a steady-state workload that requires an instance to be up most of the time, the client may reserve a High Uptime Ratio instance and potentially pay an even lower hourly usage fee, although in some embodiments the hourly fee may be charged for the entire duration of the reservation, regardless of the actual number of hours of use, in accordance with pricing policy. An option for Medium Uptime Ratio instances, with a corresponding pricing policy, may be supported in some embodiments as well, where the upfront costs and the per-hour costs fall between the corresponding High Uptime Ratio and Low Uptime Ratio costs.
Compute instance configurations may also include compute instances with a general or specific purpose, such as computational workloads for compute intensive applications (e.g., high-traffic web applications, ad serving, batch processing, video encoding, distributed analytics, high-energy physics, genome analysis, and computational fluid dynamics), graphics intensive workloads (e.g., game streaming, 3D application streaming, server-side graphics workloads, rendering, financial modeling, and engineering design), memory intensive workloads (e.g., high performance databases, distributed memory caches, in-memory analytics, genome assembly and analysis), and storage optimized workloads (e.g., data warehousing and cluster file systems). Size of compute instances, such as a particular number of virtual CPU cores, memory, cache, storage, as well as any other performance characteristic. Configurations of compute instances may also include their location, in a particular data center, availability zone, geographic, location, etc. . . . and (in the case of reserved compute instances) reservation term length. Different configurations of compute instances, as discussed below with regard to
Data processing services 220 may be various types of data processing services to perform different functions (e.g., query or other processing engines to perform functions such as anomaly detection, machine learning, data lookup, or any other type of data processing operation). For example, in at least some embodiments, data processing services 230 may include a map reduce service that creates clusters of processing nodes that implement map reduce functionality over data stored in one of data storage services 240. Various other distributed processing architectures and techniques may be implemented by data processing services 230 (e.g., grid computing, sharding, distributed hashing, etc.). Note that in some embodiments, data processing operations may be implemented as part of data storage service(s) 230 (e.g., query engines processing requests for specified data). Data processing service(s) 230 may be clients of data catalog service 220 in order to obtain structural information for performing various processing operations with respect to data sets stored in data storage service(s) 230, as provisioned resources in a pool for managed query service 270.
Data catalog service 280 may provide a catalog service that ingests, locates, and identifies data and the schema of data stored on behalf of clients in provider network 200 in data storage services 230. For example, a data set stored in a non-relational format may be identified along with a container or group in an object-based data store that stores the data set along with other data objects on behalf of a same customer or client of provider network 200. In at least some embodiments, data catalog service 280 may direct the transformation of data ingested in one data format into another data format. For example, data may be ingested into data storage service 230 as single file or semi-structured set of data (e.g., JavaScript Object Notation (JSON)). Data catalog service 280 may identify the data format, structure, or any other schema information of the single file or semi-structured set of data. In at least some embodiments, the data stored in another data format may be converted to a different data format as part of a background operation (e.g., to discover the data type, column types, names, delimiters of fields, and/or any other information to construct the table of semi-structured data in order to create a structured version of the data set). Data catalog service 280 may then make the schema information for data available to other services, computing devices, or resources, such as computing resources or clusters configured to process queries with respect to the data, as discussed below with regard to
Data storage service(s) 230 may implement different types of data stores for storing, accessing, and managing data on behalf of clients 250 as a network-based service that enables clients 250 to operate a data storage system in a cloud or network computing environment. For example, data storage service(s) 230 may include various types of database storage services (both relational and non-relational) for storing, querying, and updating data. Such services may be enterprise-class database systems that are highly scalable and extensible. Queries may be directed to a database in data storage service(s) 230 that is distributed across multiple physical resources, and the database system may be scaled up or down on an as needed basis. The database system may work effectively with database schemas of various types and/or organizations, in different embodiments. In some embodiments, clients/subscribers may submit queries in a number of ways, e.g., interactively via an SQL interface to the database system. In other embodiments, external applications and programs may submit queries using Open Database Connectivity (ODBC) and/or Java Database Connectivity (JDBC) driver interfaces to the database system.
One data storage service 230 may be implemented as a centralized data store so that other data storage services may access data stored in the centralized data store for processing and or storing within the other data storage services, in some embodiments. A may provide storage and access to various kinds of object or file data stores for putting, updating, and getting various types, sizes, or collections of data objects or files. Such data storage service(s) 230 may be accessed via programmatic interfaces (e.g., APIs) or graphical user interfaces. A centralized data store may provide virtual block-based storage for maintaining data as part of data volumes that can be mounted or accessed similar to local block-based storage devices (e.g., hard disk drives, solid state drives, etc.) and may be accessed utilizing block-based data storage protocols or interfaces, such as internet small computer interface (iSCSI).
In at least some embodiments, one of data storage service(s) 230 may be a data warehouse service that utilizes a centralized data store implemented as part of another data storage service 230. A data warehouse service as may offer clients a variety of different data management services, according to their various needs. In some cases, clients may wish to store and maintain large of amounts data, such as sales records marketing, management reporting, business process management, budget forecasting, financial reporting, website analytics, or many other types or kinds of data. A client's use for the data may also affect the configuration of the data management system used to store the data. For instance, for certain types of data analysis and other operations, such as those that aggregate large sets of data from small numbers of columns within each row, a columnar database table may provide more efficient performance. In other words, column information from database tables may be stored into data blocks on disk, rather than storing entire rows of columns in each data block (as in traditional database schemes).
Managed query service 270, as discussed below in more detail with regard to
Generally speaking, clients 250 may encompass any type of client configurable to submit network-based requests to provider network 200 via network 260, including requests for storage services (e.g., a request to create, read, write, obtain, or modify data in data storage service(s) 240, etc.) or managed query service 270 (e.g., a request to query data in a data set stored in data storage service(s) 230). For example, a given client 250 may include a suitable version of a web browser, or may include a plug-in module or other type of code module that may execute as an extension to or within an execution environment provided by a web browser. Alternatively, a client 250 may encompass an application such as a database application (or user interface thereof), a media application, an office application or any other application that may make use of storage resources in data storage service(s) 240 to store and/or access the data to implement various applications. In some embodiments, such an application may include sufficient protocol support (e.g., for a suitable version of Hypertext Transfer Protocol (HTTP)) for generating and processing network-based services requests without necessarily implementing full browser support for all types of network-based data. That is, client 250 may be an application may interact directly with provider network 200. In some embodiments, client 250 may generate network-based services requests according to a Representational State Transfer (REST)-style network-based services architecture, a document- or message-based network-based services architecture, or another suitable network-based services architecture.
In some embodiments, a client 250 may provide access to provider network 200 to other applications in a manner that is transparent to those applications. For example, client 250 may integrate with an operating system or file system to provide storage on one of data storage service(s) 240 (e.g., a block-based storage service). However, the operating system or file system may present a different storage interface to applications, such as a conventional file system hierarchy of files, directories and/or folders. In such an embodiment, applications may not need to be modified to make use of the storage system service model. Instead, the details of interfacing to the data storage service(s) 240 may be coordinated by client 250 and the operating system or file system on behalf of applications executing within the operating system environment.
Clients 250 may convey network-based services requests (e.g., access requests directed to data in data storage service(s) 240, operations, tasks, or jobs, being performed as part of data processing service(s) 230, or to interact with data catalog service 220) to and receive responses from provider network 200 via network 260. In various embodiments, network 260 may encompass any suitable combination of networking hardware and protocols necessary to establish network-based-based communications between clients 250 and provider network 200. For example, network 260 may generally encompass the various telecommunications networks and service providers that collectively implement the Internet. Network 260 may also include private networks such as local area networks (LANs) or wide area networks (WANs) as well as public or private wireless networks. For example, both a given client 250 and provider network 200 may be respectively provisioned within enterprises having their own internal networks. In such an embodiment, network 260 may include the hardware (e.g., modems, routers, switches, load balancers, proxy servers, etc.) and software (e.g., protocol stacks, accounting software, firewall/security software, etc.) necessary to establish a networking link between given client 250 and the Internet as well as between the Internet and provider network 200. It is noted that in some embodiments, clients 250 may communicate with provider network 200 using a private network rather than the public Internet.
Once a resource from a pool is provided (e.g., by receiving an identifier or other indicator of the resource to utilize), managed query service 270 may interact directly with the resource 354 in virtual compute service 210 or the resource 356 in data processing services 220 to execute queries, in various embodiments. Managed query service 270 may utilize data catalog service 280, in some embodiments to store data set schemas 352, as discussed below with regard to
Managed query service 270 may implement a managed query interface 310 to handle requests from different client interfaces, as discussed below with regard to
Managed query service 270 may implement managed query service control plane 320 to manage the operation of service resources (e.g., request dispatchers for managed query interface 310, resource planner workers for resource planner 330, or query tracker monitors for query tracker 340). Managed query service control plane 320 may direct requests to appropriate components as discussed below with regard to
Managed query service 270 may implement resource planner 330 to intelligently select available computing resources from pools for execution of queries, in some embodiments, as discussed in more detail below with regard to
In various embodiments, managed query service 270 may implement query tracker 340 in order to manage the execution of queries at compute clusters, track the status of queries, and obtain the resources for the execution of queries from resource management service 290. For example, query tracker 340 may maintain a database or other set of tracking information based on updates received from different managed query service agents implemented on provisioned computing resources (e.g., computing clusters as discussed below with regard to
Clients 400 can submit many different types of request to managed query interface 310. For example, in one embodiment, clients 400 can submit requests 450 to create, read, modify, or delete data schemas. For example, a new table schema can be submitted via a request 450. Request 450 may include a name of the data set (e.g., table), a location of the data set (e.g. an object identifier in an object storage service, such as data storage service 230, file path, uniform resource locator, or other location indicator), number of columns, column names, data types for fields or columns (e.g., string, integer, Boolean, timestamp, array, map, custom data types, or compound data types), data format (e.g., formats including, but not limited to, JSON, CSV, AVRO, ORC, PARQUET, tab delimited, comma separated, as well as custom or standard serializers/desrializers), partitions of a data set (e.g., according to time, geographic location, or other dimensions), or any other schema information for process queries with respect to data sets, in various embodiments. In at least some embodiments, request to create/read/modify/delete data set schemas may be performed using a data definition language (DDL), such as Hive Query Language (HQL). Managed query interface 310 may perform respective API calls or other requests 452 with respect to data catalog service 280, to store the schema for the data set (e.g., as part of table schemas 402). Table schemas 402 may be stored in different formats (e.g., Apache Hive). Note, in other embodiments, managed query service 270 may implement its own metadata store.
Clients 400 may also send queries 460 and query status 470 requests to managed query interface 310 which may direct those requests 460 and 470 to managed query service control plane 320, in various embodiments, as discussed below with regard to
Client(s) 400 may also submit requests for query history 480 or other account related query information (e.g., favorite or common queries) which managed query. In some embodiments, client(s) 400 may programmatically trigger the performance of past queries by sending a request to execute a saved query 490, which managed query service control plane 320 may look-up and execute. For example, execute saved query request may include a pointer or other identifier to a query stored or saved for a particular user account or client. Managed query service control plane 320 may then access that user query store to retrieve and execute the query (according to techniques discussed below with regard to
Managed query agent 512 may get schema 540 for the data sets(s) 520 from data catalog service 280, which may return the appropriate schema 542. Provisioned cluster 510 can then generate a query execution plan and execute the query 544 with respect to data set(s) 520 according to the query plan. Managed query agent 512 may send query status 546 to query tracker 340 which may report query status 548 in response to get query status 546 request, sending a response 550 indicating the query status 550. Provisioned cluster 510 may store the query results 552 in a result store 522 (which may be a data storage service 230). Managed query service control plane 320 may receive q request to get a query results 554 and get query results 556 from results store 522 and provide the query results 558 in response, in some embodiments.
Different types of computing resources may be provisioned and configured in resource pools, in some embodiments. Single-node clusters or multi-node compute clusters may be one example of a type of computing resource provisioned and configured in resource pools by resource management service 290 to service queries for managed query service 270.
Leader node 620 may implement query engine 624 to execute queries, such as query 602 which may be received via managed query agent 622 as query 603. For instance, managed query agent may implement a programmatic interface for query tracker to submit queries (as discussed above in
Query model(s) 720 can be generated using many different sources of information. For example, as illustrated in
Query model(s) 720 may also be generated using execution data 742 received from the execution of queries at computing resources of different configurations, in some embodiments. For example, the memory consumed, processing capacity consumed, number of slots, nodes, or other execution units consumed/utilized, execution plans for prior queries (e.g., including the various types of operations selected by the plan to perform the prior queries), the total execution time to perform the query, a resource or determined cost (e.g., in terms of work or cost units, such as IOPS or monetary units), failure or success indications, failure types (e.g., including error codes or exceptions generated).
Updates to query model(s) 754 may be periodically performed (e.g., daily or weekly) or in response to trigger events (e.g., number of queries processed since last update, number of new queries processed since last update, new set of execution data 742 or table metadata 754 received etc.), in some embodiments. Query model training 750 may apply different types of machine learning techniques to generate and update query model(s) 720. For instance, the information related to a prior query (e.g., execution metadata, including the query execution plan, execution costs, etc., and table metadata, including number of rows in access tables) may be used to generate feature vectors that create a feature space for performing comparisons with newly received queries, in one embodiment. Feature extraction or selection techniques may be implemented to determine which data (e.g., what kinds of table metadata or execution data) are more determinative for different performance outcomes for a query in order to adjust the query model(s), in some embodiments. Note that although supervised learning techniques are described above, in some embodiments, query model(s) 720 may be generated using unsupervised learning techniques.
Query model(s) 720 may be applied to received queries in order to determine a resource configuration, such as a query engine configuration for a received query. For example, query 702 may be received at resource planner 330. In some embodiments, resource planner 330 may implement different query plan generator(s) for query engine(s) 710. Different types of query engines may be implemented to execute queries, and thus, plan generators for the different possible query engines may be implemented in order to generate an execution plan for each query engine. For example, a query engine utilizing the Presto framework may generate a different query execution plan than a query engine utilizing the Apache Spark framework. Similarly, different query model(s) 720 may be maintained for the different query engines, in some embodiments. Query plans may include various kinds operations that may be performed to execute a query, such as different types of join operations (e.g., hash joins, distributed joins, etc.), different types of scan operations, aggregation operations, predicate filters, and the like.
The query and query execution plans may be provided 712 and evaluated using query model(s) 720. For example, a feature vector for the query based on the query execution plan and execution limitation 702 may be generated so that the feature vector can then be compared or classified (e.g., using a linear function that assigns a score for each possible configuration), in one embodiment. Scores, or other configuration information, may be determined for each query model in embodiments implementing multiple query model(s). The resulting classifications may include a number of nodes, slots, containers, or other components for a computing resource (e.g., in a cluster) as well as the configuration of (e.g., settings enabled) for a query engine. For example, different query engines may provide many different kinds of configuration or optimization settings that can be enabled or disabled, in some embodiments. As part of applying the query model(s) 720, for instance, different configurations of a Presto framework may be determined (e.g., configurations that enable or disable features such as memory configuration properties, log file location properties, worker configuration properties, coordinator configuration properties, catalog properties, Java Virtual Machine (JVM) properties, optimizer properties, node scheduler properties, exchange properties, distributed joins, distributed index joins, redistributed writes, session properties, etc.).
Application of query model(s) 720 may classify or otherwise indicate a resource configuration (e.g., query engine configuration(s) 722 for each query model that satisfies the execution limitation 704. Execution limitations 702 may be time limitations (e.g., query execution time limits), cost-defined limitations (e.g., number of resource units or cost units consumed to execute the query, service level agreements (SLAs), performance models, cost models, or any other limitation that may be placed on the execution of the query (e.g. a limitation excluding certain query execution engines from consideration). In some embodiments, resource selection 730 may first select one query engine configuration to use (e.g., based on confidence scores or other values that indicate the strength of the classification of the query in the different query model(s) 720 that satisfies the execution limitation 704 if multiple query model(s) are used.
Resource selection 730 may obtain a set (e.g., a snapshot of) available resource(s) 734 for executing the query from resource manager service. Note that the available resource(s) 734 may be provided to query selection engine 730 prior to the receipt of query 702, in some embodiments. Query selection engine 730 may then compare the available resource(s) 734 with the query engine configuration(s) 722 to select the resource to execute the query. For example, resource selection 730 may compare the number nodes in a cluster to see if the number of nodes meets or exceeds the number of nodes identified in query engine configuration(s) 722, in one embodiment. Similarly, query selection engine 730 may compare the configuration(s) of the different resources, such as the engine type (e.g., Presto, Apache Spark, etc.) and the configuration of the engine (e.g., which properties are enabled or disabled on the engine). In some embodiments, resource selection may select a candidate list or set of query engines that meet or exceed the query engine configuration(s) 722. For example, if the query engine configuration 722 identifies a cluster of 10 nodes, then 5 clusters with 10 or greater nodes may be identified as a candidate list of resources. Resource selection 730 may then provide a selected resource 732 to query tracker 340 in order to execute the query. If the selected resource is no longer available (e.g., due to failure or having already been assigned a query in the time between indicated as available 734 and selected 732), then another resource from the candidate set may be selected 732 (or the candidate set may be provided to query tracker 340, in some embodiments, for retry until assignment).
In some embodiments, query concurrency patterns 744 may alter or modify the selection of resources. For instance, as noted above a same client, user, customer, or submitter may submit multiple queries which may be routed to the same resource so that the queries can be executed without recycling or scrubbing the resource (as the queries would not be executed on resources for which the submitter of the query had not right to access the data or results of the other queries executing at that resource). Query concurrency patterns 744 may indicate the likelihood that a same submitter will submit another query in addition to query 704, as well as the type of query the subsequent query may be, in some embodiments. In some embodiments, resource selection 730 may select a query engine that can include capacity to accommodate both the received query within the execution limitation 704 as well as the likely subsequent query (which may also have an execution limitation). In this way, resource selection 730 may still allow for queries to be co-located on resources executing queries from the same submitter.
In at least some embodiments, resource planner 330 may implement resource provisioning model 740. Resource provisioning model 740 may be a model that is generated from execution data 742 for previously executed queries to determine or recommend resources (including the configuration of the resources) for provisioning, in some embodiments. For example, resource provisioning model 740 may be generated from various unsupervised learning techniques, such as clustering analysis, dimensionality reduction, and other such techniques to determine which resource configurations (e.g., query engine types, query engine configurations, number of nodes, slots, or units) are likely to be utilized by queries (e.g., according to the techniques for selection resources at resource selection 730). For example, clustering analysis may be performed to identify that ninety percent of clusters have query engine type A, configuration settings B set, and have C number of nodes. Thus, resource provisioning model 746 may provide a recommendation 746 to provision more resources with query engine type A, configuration settings B set, and have C number of nodes. In some embodiments, resource provisioning model 740 may provide general provisioning recommendations according to a set of classifications (e.g., small clusters, medium clusters, larger clusters with engine type A, and configuration settings B set). In at least some embodiments, multiple resource provisioning models 740 may be implemented (e.g., to predict the demand of different types of query engines for different times when making provisioning recommendations).
As indicated at 862, resource management service 290 may automatically (or in response to requests (not illustrated)), commission or decommission pool(s) of clusters 810. For example in some embodiments, resource management service 290 may perform techniques that select the number and size of computing clusters 820 for the warm cluster pool 810. The number and size of the computing clusters 820 in the warm cluster pool 810 can be determined based upon a variety of factors including, but not limited to, historical and/or expected volumes of query requests, the price of the computing resources utilized to implement the computing clusters 820, and/or other factors or considerations, in some embodiments.
Once the number and size of computing clusters 820 has been determined, the computing clusters 820 may be instantiated, such as through the use of an on-demand computing service, or virtual compute service or data processing service as discussed above in
Instantiated and configured computing clusters 820 that are available for use by the managed query service 270 are added to the warm cluster pool 810, in some embodiments. A determination can be made as to whether the number or size of the computing clusters 820 in the warm cluster pool needs is to be adjusted, in various embodiments. The performance of the computing clusters 820 in the warm cluster pool 810 can be monitored based on metric(s) 890 received from the cluster pool. The number of computing clusters 820 assigned to the warm cluster pool 810 and the size of each computing cluster 820 (i.e. the number of host computers in each computing cluster 820) in the warm cluster pool 810 can then be adjusted. Such techniques can be repeatedly performed in order to continually optimize the number and size of the computing clusters 820 in the warm cluster pool 810. Configurations of clusters for a resource pool or a new pool may be provided as provisioning recommendations (as discussed above with regard to
As indicated at 880, in some embodiments, resource management service 270 may scrub clusters(s) 880, (e.g., as a result of the lease state transitioning to expired or terminated) by causing the cluster to perform operations (e.g., a reboot, disk wipe, memory purge/dump, etc.) so that the cluster no longer retains client data and is ready to process another query. For example, resource management service 290 may determine whether a computing cluster 820 is inactive (e.g. the computing cluster 820 has not received a query in a predetermined amount of time). If resource management service 290 determines that the computing cluster 820 is inactive, then the computing cluster 820 may be disassociated from the submitter of the query. The computing cluster 820 may then be “scrubbed,” such as by removing data associated with the submitter of the queries from memory (e.g. main memory or a cache) or mass storage device (e.g. disk or solid state storage device) utilized by the host computers in the computing cluster 820. The computing cluster 820 may then be returned to the warm cluster pool 810 for use in processing other queries. In some embodiments, some clusters that are inactive might not be disassociated from certain users in certain scenarios. In these scenarios, the user may have a dedicated warm pool of clusters 810 available for their use.
As indicated at 860, in some embodiments, resource management service 290 may receive requests to configure resources or a pool of resources. For example, a request to configure a pool of resources may identify a type or size of cluster, a processing engine, machine image, or software to execute for individual clusters in the pool. In some embodiments, the request may indicate a maximum number of resources in the pool, a minimum number of idle resources in the pool, and a maximum number of idle resources in the pool. As indicated at 870, resource management service may receive a request to configure or specify a pool modification event for a pool, in some embodiments. For example, the pool modification event may be defined according to one or more criteria, such as the minimum number of idle resources, maximum number of idle resources, average job execution time thresholds, pool or resource lifecycle/state conditions, or any other set of one or more criteria that may be evaluated to detect a pool modification event.
Although
As indicated at 910, a first query directed to data set(s) may be received, in various embodiments. The first query may be received that is directed to data set(s) separately stored in remote data stores, in various embodiments. For example, a query may be received via the various types of interfaces described above with regard to
As indicated at 920, computing resource(s) may be selected to execute the first query based, at least in part, on how the computing resource(s) are configured compared to computing resources used in performing prior queries, in some embodiments. For example, a historical query model (e.g., such as a model generated according to machine learning techniques) may be maintained that models the performance of queries with different characteristics based on different execution outcomes (e.g., time to complete, cost to complete, resources consumed to complete, probability of failure or timeout, etc.). The historical query model may then be applied, in some embodiments to the query, by comparing features of the query with respect to the features of queries identified in the model, as discussed in detail below with regard to
As indicated at 930, the first query may be performed at the selected computing resource(s) with respect to the data set(s), in various embodiments. For example, the resource request may be routed to the selected resource in some embodiments. A request to initiate or being processing at the selected computing resource(s) may be performed, in some embodiments, according to an API request or the first query may be initiated by transmitting the query in its original format to the computing resources for execution.
As indicated at 1020, the first query may be evaluated with respect to resource configurations for executing the first query, in some embodiments. For example, a determination may be made as to whether the first query is the same or similar to prior query (e.g., by comparing query language or data sources). If so, then the resource configuration(s) of the same or prior queries may be evaluated according to a desired performance outcome or limitation (as discussed above). If no similar queries can be found, then a default computing resource configuration may be selected or further types of similarity analysis or classification may be performed (e.g., using a historical query model as discussed above and below). As indicated at 1030, a computing resource may be selected to execute the query from a plurality of differently configured computing resources that execute queries, based at least in part, on the resource configuration, in some embodiments. The plurality of computing resources may be pre-configured according to query engines of different types, different configuration settings, and/or different sizes (e.g., number of nodes or slots in a cluster). The determined resource configuration may be compared with the configured computing resources to determine a best match or other similarity score for the determine resource configurations and the configuration of resources. In some embodiments, selection of a computing resource may include selection of a pool of resources (of a same configuration), so that the actual resource selected is determined from the pool.
As indicated at 1040, the first query may be routed to the selected computing resource(s) for execution, in some embodiments. As indicated at 1050, a result for the query generated at the selected computing resource(s) may be provided, in some embodiments. For example, the results can be sent to a destination or location specified for the query results (e.g., in a client request), in one embodiment. The results may be streamed back or aggregated (e.g., in a data store, like data storage service 230) and provided as a batch (or batches, such as paginated results) via a same interface (e.g., programmatic, graphical, driver, console, etc.) that received the query.
As indicated at 1060, a query execution model according to execution data for executing the query at the resources, in some embodiments. For example, the memory consumed, processing capacity consumed, number of slots, nodes, or other execution units consumed/utilized, execution plans for prior queries (e.g., including the various types of operations selected by the plan to perform the prior queries), the total execution time to perform the query, a resource or determined cost (e.g., in terms of work or cost units, such as IOPS or monetary units), failure or success indications, and/or failure types (e.g., including error codes or exceptions generated) may be included as part of execution data. A supervised learning technique may take the execution data as part of the training set, mapping the performance or results of the query's execution to the features of the query (e.g., query execution plan, source data sets, etc.) in order to train the query execution model to achieve a similar outcome for the query if a similar query is received.
As indicated at 1120, query execution plan(s) for the first query for query engine(s) may be generated, in some embodiments. For example, query plan generation techniques that parse a SQL query, identify the SQL operations, and data identified in the query, determine the operations to perform in order to accomplish the query, and assemble the operations in multiple candidate execution plans (e.g., based on table statistics or other metadata for queried data sets) may be performed. Then the candidate execution plan based on a lowest cost to execute the plan may be selected so that optimized query execution plans are generated for the query, in some embodiments. If multiple query engines may be considered to execute the query, then multiple query execution plans may be generated, in some embodiments, one for each type of query engine.
As indicated at 1130, historical query execution model(s) may be applied to the first query and the query execution plan(s) for one of the query execution(s) for the query according to an execution limitation for the query, in some embodiments. For example, feature vectors or other representations of the query and query plan(s) may be generated in order to analyze the received query with respect to the historical query model. A classification or regression function determined from the historical query execution model(s) may be applied, in some embodiments to generate similarity scores between a query and different classes of queries that may be described in the historical query execution model (e.g., short-running, long-running, compute intensive, memory intensive, DDL query, or DML query). The execution limitation may be applied to determine which classifications do (or do not) satisfy the execution limitation (e.g., a time limit for performing the query or cost limit for performing the query), in some embodiments. Alternatively, different classifications may be mapped to different execution limitations (e.g., classification A satisfies limit A, classification B satisfies limit B, etc.).
As indicated at 1140, a computing resource may be selected from available computing resources according to the similarity of the available computing resources to the determined configuration. For example, a set (e.g., a snapshot of) available computing resources for executing the query may be obtained (e.g., by examining availability data, resource pool data, pinging resources to see if they are available, or by requesting the set of available computing resources from a service like resource management service 290 in
If a provisioning event is detected, then a resource provisioning model generated from prior executed queries to determine computing resource(s) to provision may be evaluated, as indicated at 1230, in various embodiments. For example, the resource provisioning model may be evaluated with respect to one type of resource configuration criteria (e.g., number of nodes in a cluster) in order to determine other resource configuration criteria (e.g., engine type and/or engine configuration settings). Other variables such as the time of day, overall state of resource pool(s), or other information that may alter a provisioning recommendation (e.g., including excluding some resource(s) from configuration) may be used to evaluate the resource provisioning model. In some embodiments, resources to be provisioned may be determined according to general classifications, such as small, medium, or large clusters, or may be determined with a specific number of nodes, engine type and engine configuration settings. In some embodiments, a time-based analysis of the execution of prior queries and resource configurations may be performed (e.g., examining demand and resource configuration as a time series) to predict the demand, and thus number and configuration of resources to include the pool based on a time or time period associated with the provisioning event. For instance, a provisioning event to launch a resource pool at 6:00 PM EST may evaluate the demand for resources starting at 6:00 PM EST, as well as the configuration of the resources used to execute the queries received in order to provision a number of resources that can satisfy a predicted demand for the pool starting at 6:00 PM EST.
The determined resource(s) may then be provisioned, as indicated at 1230, in some embodiments. For example, requests to other network-based systems or services to launch, create, instantiate, or configure new resources according to the configuration of the determined computing resources.
In some embodiments, pools of computing resource(s) configured for query execution may be monitored, in various embodiments, to provisioning individual resources within a pool of computing resources. For example, as resources are leased, assigned, or otherwise allocated to execute queries, the number of available resources in the pool may decrease. Similarly, resource failures or resource expirations may reduce the number of available resources. If, for example, the number of available resources falls below a maintenance threshold, then a provisioning event may be detected for an existing pool. Similar techniques to those described above may be performed to provide a recommendation as to the number and/or configuration of resources to provision for the existing pool.
The methods described herein may in various embodiments be implemented by any combination of hardware and software. For example, in one embodiment, the methods may be implemented by a computer system (e.g., a computer system as in
Each type of computing resource provided by the service provider network 200 can be general-purpose or can be available in a number of specific configurations. For example, data processing resources can be available as physical computers or VM instances in a number of different configurations. The VM instances can execute applications, including web servers, application servers, media servers, database servers, some or all of the services described above, and/or other types of programs. The VM instances can also be configured into computing clusters in the manner described above. Data storage resources can include file storage devices, block storage devices, and the like. The service provider network 200 can also provide other types of computing resources not mentioned specifically herein.
The computing resources provided by the service provider network maybe implemented, in some embodiments, by one or more data centers 1304A-1304N (which might be referred to herein singularly as “a data center 1304” or in the plural as “the data centers 1304”). The data centers 1304 are facilities utilized to house and operate computer systems and associated components. The data centers 1304 typically include redundant and backup power, communications, cooling, and security systems. The data centers 1304 can also be located in geographically disparate locations. One illustrative configuration for a data center 1304 that can be utilized to implement the technologies disclosed herein will be described below with regard to
The customers and other users of the service provider network 200 can access the computing resources provided by the service provider network 200 over a network 1302, which can be a wide area communication network (“WAN”), such as the Internet, an intranet or an Internet service provider (“ISP”) network or a combination of such networks. For example, and without limitation, a computing device 1300 operated by a customer or other user of the service provider network 200 can be utilized to access the service provider network 200 by way of the network 1302. It should be appreciated that a local-area network (“LAN”), the Internet, or any other networking topology known in the art that connects the data centers 1304 to remote customers and other users can be utilized. It should also be appreciated that combinations of such networks can also be utilized.
The server computers 1402 can be standard tower, rack-mount, or blade server computers configured appropriately for providing the computing resources described herein (illustrated in
The data center 1304 shown in
In the example data center 1304 shown in
Embodiments of a managed query execution as described herein may be executed on one or more computer systems, which may interact with various other devices. One such computer system is illustrated by
In the illustrated embodiment, computer system 2000 includes one or more processors 2010 coupled to a system memory 2020 via an input/output (I/O) interface 2030. Computer system 2000 further includes a network interface 2040 coupled to I/O interface 2030, and one or more input/output devices 2050, such as cursor control device 2060, keyboard 2070, and display(s) 2080. Display(s) 2080 may include standard computer monitor(s) and/or other display systems, technologies or devices. In at least some implementations, the input/output devices 2050 may also include a touch- or multi-touch enabled device such as a pad or tablet via which a user enters input via a stylus-type device and/or one or more digits. In some embodiments, it is contemplated that embodiments may be implemented using a single instance of computer system 2000, while in other embodiments multiple such systems, or multiple nodes making up computer system 2000, may host different portions or instances of embodiments. For example, in one embodiment some elements may be implemented via one or more nodes of computer system 2000 that are distinct from those nodes implementing other elements.
In various embodiments, computer system 2000 may be a uniprocessor system including one processor 2010, or a multiprocessor system including several processors 2010 (e.g., two, four, eight, or another suitable number). Processors 2010 may be any suitable processor capable of executing instructions. For example, in various embodiments, processors 2010 may be general-purpose or embedded processors implementing any of a variety of instruction set architectures (ISAs), such as the x86, PowerPC, SPARC, or MIPS ISAs, or any other suitable ISA. In multiprocessor systems, each of processors 2010 may commonly, but not necessarily, implement the same ISA.
In some embodiments, at least one processor 2010 may be a graphics processing unit. A graphics processing unit or GPU may be considered a dedicated graphics-rendering device for a personal computer, workstation, game console or other computing or electronic device. Modern GPUs may be very efficient at manipulating and displaying computer graphics, and their highly parallel structure may make them more effective than typical CPUs for a range of complex graphical algorithms. For example, a graphics processor may implement a number of graphics primitive operations in a way that makes executing them much faster than drawing directly to the screen with a host central processing unit (CPU). In various embodiments, graphics rendering may, at least in part, be implemented by program instructions configured for execution on one of, or parallel execution on two or more of, such GPUs. The GPU(s) may implement one or more application programmer interfaces (APIs) that permit programmers to invoke the functionality of the GPU(s). Suitable GPUs may be commercially available from vendors such as NVIDIA Corporation, ATI Technologies (AMD), and others.
System memory 2020 may store program instructions and/or data accessible by processor 2010. In various embodiments, system memory 2020 may be implemented using any suitable memory technology, such as static random access memory (SRAM), synchronous dynamic RAM (SDRAM), nonvolatile/Flash-type memory, or any other type of memory. In the illustrated embodiment, program instructions and data implementing desired functions, such as those described above are shown stored within system memory 2020 as program instructions 2025 and data storage 2035, respectively. In other embodiments, program instructions and/or data may be received, sent or stored upon different types of computer-accessible media or on similar media separate from system memory 2020 or computer system 2000. Generally speaking, a non-transitory, computer-readable storage medium may include storage media or memory media such as magnetic or optical media, e.g., disk or CD/DVD-ROM coupled to computer system 2000 via I/O interface 2030. Program instructions and data stored via a computer-readable medium may be transmitted by transmission media or signals such as electrical, electromagnetic, or digital signals, which may be conveyed via a communication medium such as a network and/or a wireless link, such as may be implemented via network interface 2040.
In one embodiment, I/O interface 2030 may coordinate I/O traffic between processor 2010, system memory 2020, and any peripheral devices in the device, including network interface 2040 or other peripheral interfaces, such as input/output devices 2050. In some embodiments, I/O interface 2030 may perform any necessary protocol, timing or other data transformations to convert data signals from one component (e.g., system memory 2020) into a format suitable for use by another component (e.g., processor 2010). In some embodiments, I/O interface 2030 may include support for devices attached through various types of peripheral buses, such as a variant of the Peripheral Component Interconnect (PCI) bus standard or the Universal Serial Bus (USB) standard, for example. In some embodiments, the function of I/O interface 2030 may be split into two or more separate components, such as a north bridge and a south bridge, for example. In addition, in some embodiments some or all of the functionality of I/O interface 2030, such as an interface to system memory 2020, may be incorporated directly into processor 2010.
Network interface 2040 may allow data to be exchanged between computer system 2000 and other devices attached to a network, such as other computer systems, or between nodes of computer system 2000. In various embodiments, network interface 2040 may support communication via wired or wireless general data networks, such as any suitable type of Ethernet network, for example; via telecommunications/telephony networks such as analog voice networks or digital fiber communications networks; via storage area networks such as Fibre Channel SANs, or via any other suitable type of network and/or protocol.
Input/output devices 2050 may, in some embodiments, include one or more display terminals, keyboards, keypads, touchpads, scanning devices, voice or optical recognition devices, or any other devices suitable for entering or retrieving data by one or more computer system 2000. Multiple input/output devices 2050 may be present in computer system 2000 or may be distributed on various nodes of computer system 2000. In some embodiments, similar input/output devices may be separate from computer system 2000 and may interact with one or more nodes of computer system 2000 through a wired or wireless connection, such as over network interface 2040.
As shown in
Those skilled in the art will appreciate that computer system 2000 is merely illustrative and is not intended to limit the scope of the techniques as described herein. In particular, the computer system and devices may include any combination of hardware or software that can perform the indicated functions, including a computer, personal computer system, desktop computer, laptop, notebook, or netbook computer, mainframe computer system, handheld computer, workstation, network computer, a camera, a set top box, a mobile device, network device, internet appliance, PDA, wireless phones, pagers, a consumer device, video game console, handheld video game device, application server, storage device, a peripheral device such as a switch, modem, router, or in general any type of computing or electronic device. Computer system 2000 may also be connected to other devices that are not illustrated, or instead may operate as a stand-alone system. In addition, the functionality provided by the illustrated components may in some embodiments be combined in fewer components or distributed in additional components. Similarly, in some embodiments, the functionality of some of the illustrated components may not be provided and/or other additional functionality may be available.
Those skilled in the art will also appreciate that, while various items are illustrated as being stored in memory or on storage while being used, these items or portions of them may be transferred between memory and other storage devices for purposes of memory management and data integrity. Alternatively, in other embodiments some or all of the software components may execute in memory on another device and communicate with the illustrated computer system via inter-computer communication. Some or all of the system components or data structures may also be stored (e.g., as instructions or structured data) on a computer-accessible medium or a portable article to be read by an appropriate drive, various examples of which are described above. In some embodiments, instructions stored on a non-transitory, computer-accessible medium separate from computer system 2000 may be transmitted to computer system 2000 via transmission media or signals such as electrical, electromagnetic, or digital signals, conveyed via a communication medium such as a network and/or a wireless link. Various embodiments may further include receiving, sending or storing instructions and/or data implemented in accordance with the foregoing description upon a computer-accessible medium. Accordingly, the present invention may be practiced with other computer system configurations.
It is noted that any of the distributed system embodiments described herein, or any of their components, may be implemented as one or more web services. For example, leader nodes within a data warehouse system may present data storage services and/or database services to clients as network-based services. In some embodiments, a network-based service may be implemented by a software and/or hardware system designed to support interoperable machine-to-machine interaction over a network. A network-based service may have an interface described in a machine-processable format, such as the Web Services Description Language (WSDL). Other systems may interact with the web service in a manner prescribed by the description of the network-based service's interface. For example, the network-based service may define various operations that other systems may invoke, and may define a particular application programming interface (API) to which other systems may be expected to conform when requesting the various operations.
In various embodiments, a network-based service may be requested or invoked through the use of a message that includes parameters and/or data associated with the network-based services request. Such a message may be formatted according to a particular markup language such as Extensible Markup Language (XML), and/or may be encapsulated using a protocol such as Simple Object Access Protocol (SOAP). To perform a web services request, a network-based services client may assemble a message including the request and convey the message to an addressable endpoint (e.g., a Uniform Resource Locator (URL)) corresponding to the web service, using an Internet-based application layer transfer protocol such as Hypertext Transfer Protocol (HTTP).
In some embodiments, web services may be implemented using Representational State Transfer (“RESTful”) techniques rather than message-based techniques. For example, a web service implemented according to a RESTful technique may be invoked through parameters included within an HTTP method such as PUT, GET, or DELETE, rather than encapsulated within a SOAP message.
The various methods as illustrated in the FIGS. and described herein represent example embodiments of methods. The methods may be implemented in software, hardware, or a combination thereof. The order of method may be changed, and various elements may be added, reordered, combined, omitted, modified, etc.
Various modifications and changes may be made as would be obvious to a person skilled in the art having the benefit of this disclosure. It is intended that the invention embrace all such modifications and changes and, accordingly, the above description to be regarded in an illustrative rather than a restrictive sense.
This application is a continuation of U.S. patent application Ser. No. 15/470,843, filed Mar. 27, 2017, which claims benefit of priority to U.S. Provisional Application Ser. No. 62/382,477, file Sep. 1, 2016, and which are incorporated herein by reference in their entirety.
Number | Name | Date | Kind |
---|---|---|---|
6314447 | Lea et al. | Nov 2001 | B1 |
6366915 | Rubert et al. | Apr 2002 | B1 |
6473750 | Petculescu et al. | Oct 2002 | B1 |
6859926 | Brenner et al. | Feb 2005 | B1 |
7243093 | Cragun et al. | Jul 2007 | B2 |
7406461 | Chapman et al. | Jul 2008 | B1 |
8429096 | Soundararajan et al. | Apr 2013 | B1 |
8782075 | Zane et al. | Jul 2014 | B2 |
9208032 | McAlister et al. | Dec 2015 | B1 |
10614066 | Gawande et al. | Apr 2020 | B2 |
10762086 | Wu et al. | Sep 2020 | B2 |
20040205759 | Oka | Oct 2004 | A1 |
20080201459 | Vul et al. | Aug 2008 | A1 |
20100094862 | Bent et al. | Apr 2010 | A1 |
20100094892 | Bent et al. | Apr 2010 | A1 |
20120179644 | Miranker | Jul 2012 | A1 |
20130024442 | Santosuosso et al. | Jan 2013 | A1 |
20130160014 | Watanabe et al. | Jun 2013 | A1 |
20130262638 | Kumarasamy | Oct 2013 | A1 |
20140032535 | Singla | Jan 2014 | A1 |
20140074540 | Evans et al. | Mar 2014 | A1 |
20140149355 | Gupta et al. | May 2014 | A1 |
20140156632 | Yu et al. | Jun 2014 | A1 |
20140229221 | Shih et al. | Aug 2014 | A1 |
20140280076 | Sumizawa | Sep 2014 | A1 |
20150040180 | Jacobson et al. | Feb 2015 | A1 |
20150149501 | Prakash et al. | May 2015 | A1 |
20150234922 | Dageville et al. | Aug 2015 | A1 |
20160373478 | Doubleday | Dec 2016 | A1 |
20170316078 | Funke et al. | Nov 2017 | A1 |
20180039674 | Seyvet et al. | Feb 2018 | A1 |
20180060393 | Kalathuru et al. | Mar 2018 | A1 |
20180088566 | Billi-Duran et al. | Mar 2018 | A1 |
Number | Date | Country |
---|---|---|
2778967 | Sep 2014 | EP |
2012058815 | Mar 2012 | JP |
2015529366 | Oct 2015 | JP |
2014039919 | Mar 2014 | WO |
Entry |
---|
Anonymous, “optimization—Are SQL Execution Plans based on Schema or Data or both?”, Stack Overflow.com, Retrieved from the Internet: URL:https://web.archive.org/web/20160708184640/https://stackoverflow.com/questions/4787205/are-sql-execution-plans-based-onschema-or-data-or-both, Jul. 8, 2016. |
Number | Date | Country | |
---|---|---|---|
20200233869 A1 | Jul 2020 | US |
Number | Date | Country | |
---|---|---|---|
62382477 | Sep 2016 | US |
Number | Date | Country | |
---|---|---|---|
Parent | 15470843 | Mar 2017 | US |
Child | 16839849 | US |