The present description relates to data storage and retrieval and, more specifically, to a technique for comparing system performance load to a realistic performance benchmark that accounts for intricacies of the workload.
Networks and distributed storage allow data and storage space to be shared between devices located anywhere a connection is available. These implementations may range from a single machine offering a shared drive over a home network to an enterprise-class cloud storage array with multiple copies of data distributed throughout the world. Larger implementations may incorporate Network Attached Storage (NAS) devices, Storage Area Network (SAN) devices, and other configurations of storage elements and controllers in order to provide data and manage its flow. Improvements in distributed storage have given rise to a cycle where applications demand increasing amounts of data delivered with reduced latency, greater reliability, and greater throughput. Building out a storage architecture to meet these expectations enables the next generation of applications, which is expected to bring even greater demand.
In order to provide storage solutions that meet a customer's needs and budget, it is not sufficient to blindly add hardware. Instead, it is increasingly beneficial to seek out and reduce bottlenecks, limitations in one aspect of a system that prevent other aspects from operating at their full potential. For example, an administrator may desire to know whether adding additional storage devices to a storage system will improve performance or whether moving data sets between devices can deliver the same benefits without adding hardware. An administrator may also want to know how much of a performance impact to expect if another application is hosted on a storage system. As another example, as costs of solid-state devices have come down, administrators are being asked to identify data that should be transitioned from magnetic hard disk drives to solid-state devices in order to maximize performance.
Accordingly, a need exists for improved performance tracking and analysis techniques capable of providing a more accurate assessment of performance under real-world conditions. In particular, systems and methods for comparing storage system performance against realistic performance benchmarks may provide a more comprehensive picture that enables administrators to reduce bottlenecks and to improve data access speeds. Thus, while existing techniques for storage system benchmarking have been generally adequate, the techniques described herein provide improved accuracy and efficiency.
The present disclosure is best understood from the following detailed description when read with the accompanying figures.
All examples and illustrative references are non-limiting and should not be used to limit the claims to specific implementations and embodiments described herein and their equivalents. For simplicity, reference numbers may be repeated between various examples. This repetition is for clarity only and does not dictate a relationship between the respective embodiments except where explicitly noted. Finally, in view of this disclosure, particular features described in relation to one aspect or embodiment may be applied to other disclosed aspects or embodiments of the disclosure, even though not specifically shown in the drawings or described in the text.
Various embodiments include systems and methods for characterizing data transactions by identifying transaction characteristics that indicate the burden on a storage system tasked with the transactions. Any suitable transaction characteristics may be considered including: randomness, queue depth, whether the transaction is a read-or-write, I/O size, seek distance, etc. In some embodiments, the optimal system performance under such a burden is determined using a performance profile database that contains measured performance of various storage components cross-referenced by transaction characteristics. The optimal system performance may be used as a benchmark to compare with the actual performance attained by the storage system. From the comparison, the remaining performance capacity (headroom) of the storage system may be determined and may be presented at a user display. In applications where the data transactions push the computing system to its limits, such information may be used to identify upgrades that would further improve performance. In applications where the computing system has additional performance headroom, such information may be used to quantify the amount of headroom and to identify additional tasks to assign to the system.
Although the scope of embodiments is not limited to any particular use case, in one example, a storage environment includes a storage system and a performance monitor that may or may not be part of the storage system. As the storage system receives and performs data transactions from various attached hosts or from applications running on the storage system, the performance monitor identifies one or more transactions attributes relating to the performance cost or burden of the current workload. The performance monitor queries a performance profile based on the transaction attributes to determine benchmark performance metrics for the storage devices and/or other elements. The performance monitor may then compare the performance of the storage system to the benchmark. Because the performance benchmark is responsive to the type of transactions currently being executed by the storage system, it provides a more accurate indication of how much performance could be wrung from the storage system if the current workload scaled than most conventional techniques can provide. This may enable administrators to better utilize current resources and to better plan system expansion. It is understood that these features and advantages are shared among the various examples herein and that no one feature or advantage is required for any particular embodiment.
While the storage system 102 and each of the hosts 104 are referred to as singular entities, a storage system 102 or host 104 may include any number of computing devices and may range from a single computing system to a system cluster of any size. Accordingly, each storage system 102 and host 104 includes at least one computing system, which in turn includes a processor such as a microcontroller or a central processing unit (CPU) operable to perform various computing instructions. The computing system may also include a memory device such as random access memory (RAM); a non-transitory computer-readable storage medium such as a magnetic hard disk drive (HDD), a solid-state drive (SSD), or an optical memory (e.g., CD-ROM, DVD, BD); a video controller such as a graphics processing unit (GPU); a communication interface such as an Ethernet interface, a Wi-Fi (IEEE 802.11 or other suitable standard) interface, or any other suitable wired or wireless communication interface; and/or a user I/O interface coupled to one or more user I/O devices such as a keyboard, mouse, pointing device, or touchscreen.
With respect to the storage system 102, the exemplary storage system 102 contains any number of storage devices 106 and responds to hosts' data transactions so that the storage devices 106 appear to be directly connected (local) to the hosts 104. In various examples, the storage devices 106 include hard disk drives (HDDs), solid state drives (SSDs), optical drives, and/or any other suitable volatile or non-volatile data storage medium. In some embodiments, the storage devices 106 are relatively homogeneous (e.g., having the same manufacturer, model, and/or configuration). However, it is also common for the storage system 102 to include a heterogeneous set of storage devices 106 that includes storage devices of different media types from different manufacturers with notably different performance.
The storage system 102 may group the storage devices 106 for speed and/or redundancy using a virtualization technique such as RAID (Redundant Array of Independent/Inexpensive Disks). The storage system may also arrange the storage devices 106 hierarchically for improved performance by including a large pool of relatively slow storage devices and one or more caches (i.e., smaller memory pools typically utilizing faster storage media). Portions of the address space are mapped to the cache so that transactions directed to mapped addresses can be serviced using the cache. Accordingly, the larger and slower memory pool is accessed less frequently and in the background. In an embodiment, a storage device includes HDDs, while an associated cache includes SSDs.
The storage system 102 also includes one or more storage controllers 108 in communication with the storage devices 106 and any respective caches. The storage controllers 108 exercise low-level control over the storage devices in order to execute (perform) data transactions on behalf of the hosts 104. The storage system 102 may also be communicatively coupled to a user display 114 for displaying diagnostic information, application output, and/or other suitable data.
With respect to the hosts 104, a host 104 includes any computing resource that is operable to exchange data with a storage system 102 by providing (initiating) data transactions to the storage system 102. In an exemplary embodiment, a host 104 includes a host bus adapter (HBA) 110 in communication with a storage controller 108 of the storage system 102. The HBA 110 provides an interface for communicating with the storage controller 108, and in that regard, may conform to any suitable hardware and/or software protocol. In various embodiments, the HBAs 110 include Serial Attached SCSI (SAS), iSCSI, InfiniBand, Fibre Channel, and/or Fibre Channel over Ethernet (FCoE) bus adapters. Other suitable protocols include SATA, eSATA, PATA, USB, and FireWire. The HBAs 110 of the hosts 104 may be coupled to the storage system 102 by a direct connection (e.g., a single wire or other point-to-point connection), a networked connection, or any combination thereof. Examples of suitable network architectures 112 include a Local Area Network (LAN), an Ethernet subnet, a PCI or PCIe subnet, a switched PCIe subnet, a Wide Area Network (WAN), a Metropolitan Area Network (MAN), the Internet, or the like. In many embodiments, a host 104 has multiple communicative links with a single storage system 102 for redundancy. The multiple links may be provided by a single HBA 110 or multiple HBAs 110 within the hosts 104. In some embodiments, the multiple links operate in parallel to increase bandwidth.
To interact with (e.g., read, write, modify, etc.) remote data, a host HBA 110 sends one or more data transactions to the respective storage system 102. Data transactions are requests to read, write, or otherwise access data stored within a data storage device such as the storage system 102, and may contain fields that encode a command, data (i.e., information read or written by an application), metadata (i.e., information used by a storage system to store, retrieve, or otherwise manipulate the data such as a physical address, a logical address, a current location, data attributes, etc.), and/or any other relevant information. The storage system 102 executes the data transactions on behalf of the hosts 104 by reading, writing, or otherwise accessing data on the storage devices 106. A storage system 102 may also execute data transactions based on applications running on the storage system 102 using the storage devices 106. For some data transactions, the storage system 102 formulates a response that may include requested data, status indicators, error messages, and/or other suitable data and provides the response to the provider of the transaction.
In order to assess system performance and determine how quickly data transactions are being satisfied, a number of metrics may be tracked as a storage system 102 responds to real-world or synthetic data transactions. Accordingly, the storage environment 100 includes a performance monitor 116, which may or may not be part of the storage system 102 executing the transactions. In embodiments in which the performance monitor 116 is a separate computing system, similar to the storage system 102, while the performance monitor 116 is referred to as a single entity, it may include any number of computing devices and/or computing systems in communication with the storage system(s) 102 to be monitored. By offloading some of the burden of performance tracking and analysis to a separate performance monitor 116, the resources of the storage system 102 can be dedicated to servicing the data transactions.
Taken alone, the raw performance metrics may not necessarily provide sufficient context to assess whether a storage system 102 is operating at its peak. It has been determined through experimentation and analysis that certain attributes of data transactions may dramatically affect the peak performance a storage system 102 can attain. For example, some transactions, by their nature, simply have a greater burden or execution cost than others. Accordingly, some of the techniques of the present disclosure use various attributes of the data transactions to determine the workload burden and, accordingly, the maximum performance that can be obtained from the storage system 102 in light of the current environment. This gives a more realistic assessment of the storage system 102 and may give an administrator a greater insight into how much headroom (additional capacity) remains, where hardware can be added to improve performance, whether to rebalance hosts 104 or applications, and whether to take other corrective actions.
Some such techniques for storage system performance analysis are described with reference to
Referring to block 202 of
As an example, in various embodiments, the performance metrics of the database 400 include average latency, peak latency, average TOPS (I/O operations per second), peak IOPS, data throughput and/or other performance metrics. The performance metrics are correlated to characteristics that account for various data transaction attributes, device attributes, and/or system attributes determined to have an effect on performance. Such characteristics may be selected because they represent the cost (i.e., the burden) of completing the transactions. In an example, the transaction characteristics include a measurement of address randomness, and the performance profile database 400 has a first performance metric value for a data transaction that is part of a sequential set of transactions (directed to adjacent or nearby memory addresses) and a second performance metric value for a data transaction that is part of a random set of transactions (directed to non-adjacent memory addresses). In some examples, the transaction characteristics include transaction seek distance, and the performance profile database 400 records a range of performance values corresponding to a range transaction seek distances. A seek distance for a transaction directed to a hard disk drive-type storage device 106 may account for both the distance an actuator moves a read/write head and the distance a platter rotates in order to align the read/write head with the data. In some examples, the transaction characteristics include the ratio of reads to writes of the last n transactions, sometimes referred to as the operational mix. In further examples, the device, system, and transaction characteristics of the performance profile database 400 include queue depth (e.g., a depth of a storage controller queue) at the time the data transaction is performed, I/O size, spindle speed, power saving mode, interface data type, interface data protocol, interface data rate, cache status, and/or other suitable attributes.
An exemplary performance profile database 400 is illustrated in
Referring to block 204 of
Referring to block 206 of
Referring to block 208 of
Referring to block 212 of
Referring to block 214 of
Referring to block 216 of
To determine the component benchmarks, the storage system 102 and/or the performance monitor 116 may query the performance profile database 400 based on a combination of the device attributes of block 204, the storage system attributes of block 206, and/or the data transaction attributes of block 212 to determine device-level performance metrics. In such embodiments, the device-level performance metrics are combined in view of the system configuration and other storage system attributes to determine the benchmark performance level for the storage system 102. For example, for independent storage devices 106 in an independent arrangement (e.g., RAID level 0), IOPS and bandwidth of each device may add linearly, while latency may be relatively unaffected. For storage devices 106 in a redundant configuration (e.g., RAID level 1), latency and IOPS for the array may reflect the slowest storage device 106.
Once the storage system benchmark for the workload has been determined, referring to block 218 of
To aid in this process, referring to block 220 of
The user interface 600 may include a number of performance panes 602 for displaying performance attributes of elements of the storage system. In the illustrated embodiments, one of the performance panes 602 corresponds to overall storage system performance, while some of the remaining performance panes 602 correspond to performance of individual storage devices 106 of the storage system 102. The user interface 600 may include a storage element selection mechanism 604 operable to receive a user selection of those elements of the storage system 102 to display in a performance pane 602. The user interface 600 may also include a metric selection mechanism 606 operable to receive a user selection of one or more metrics to display. In some embodiments, each performance pane 602 may display a different performance metric or a different set of metrics for the respective device. The selected metric(s) may be displayed using any combination of charts, graphics, and/or text, and in one example, a performance pane 602 displays a measurement of IOPS serviced by the corresponding storage system 102 component over a period of time as determined in block 214.
The user interface may also have a benchmark selection mechanism 608 operable to receive a user selection of whether to display the benchmark performance levels determined in block 216. In the illustrated embodiment, if selected, the benchmark performance level is displayed as an overlay on the actual performance metric determined in block 214.
The user interface 600 may also include one or more workload selection mechanisms 610 operable to receive a user selection specifying what portions of the workload (i.e., what transactions of the workload) to display and/or highlight. In the illustrated embodiment, the workload selection mechanisms 610 allow a user to highlight the portion of the workload produced by a given application.
Suitable selection mechanisms for the storage element selection mechanism 604, the metric selection mechanism 606, the benchmark selection mechanism 608, and/or the workload selection mechanisms 610 include checkboxes, exclusive and non-exclusive lists, radio buttons, and other suitable interface schemes. It is understood that this is merely one example of a suitable user interface 600 and other embodiments utilize other interface configurations.
Embodiments of the present disclosure can take the form of a computer program product accessible from a tangible computer-usable or computer-readable medium providing program code for use by or in connection with a computer or any instruction execution system. For the purposes of this description, a tangible computer-usable or computer-readable medium can be any apparatus that can store the program for use by or in connection with the instruction execution system, apparatus, or device. The medium can be an electronic, magnetic, optical, electromagnetic, infrared, or a semiconductor system (or apparatus or device). In some embodiments, one or more processors running in one or more of the hosts 104 and the storage system 102 execute code to implement the actions described above.
Thus, the present disclosure provides a system and method for performance tracking and analysis that accounts for the impact of a workload on a storage system's optimal performance. In some embodiments, a computing device is provided that comprises a memory containing machine readable medium comprising machine executable code having stored thereon instructions for performing a method of performance analysis; and a processor coupled to the memory. The processor configured to execute the machine executable code to identify a data transaction received by a storage system and directed to a storage device. The processor identifies an attribute of the data transaction relating to a performance cost of the data transaction and queries a performance profile based on the identified attribute to determine a benchmark performance level for the storage device. Based on the benchmark performance level for the storage device, the processor is further operable to determine a benchmark performance level for the storage system in performing the data transaction. The computing device is further operable to compare a metric of the performance of the data transaction by the storage system with the benchmark performance level to determine remaining headroom of the storage system. In some embodiments, the attribute of the data transaction includes at least one of: an I/O size of the data transaction or a seek distance of the data transaction. In some such embodiments, the attribute of the data transaction further includes at least one of: a randomness metric, a cache status, a queue depth, or whether the transaction is a read transaction or a write transaction.
In further embodiments, a method is provided that includes receiving a data transaction at a storage system. A characteristic of the data transaction relating to an execution burden of the data transaction is identified. A data structure recording benchmark performance of a storage device of the storage system is received and, using a computer system, a performance benchmark for the storage system in executing the data transaction is determined based on the identified characteristic and the received data structure. An execution of the data transaction by the storage system is assessed to determine a performance metric, and the performance metric is compared to the determined performance benchmark to determine a remaining execution capacity of the storage system. In some such embodiments, the comparing of the performance metric to the determined performance benchmark further determines a remaining execution capacity of the storage device. In some embodiments, the determining of the performance benchmark includes: identifying a redundant array of independent disks (RAID) configuration of the storage device within the storage system, and determining the performance benchmark further based on the RAID configuration of the storage device.
In yet further embodiments, an apparatus is provided comprising a non-transitory machine-readable medium having stored thereon instructions for performing a method of workload assessment. The instructions comprise machine executable code, which when executed by at least one machine, causes the machine to: characterize an execution burden of a data transaction; determine, based on a benchmark performance record and the characterized execution burden, a component benchmark for a component of a storage system performing the data transaction; determine a system benchmark for the storage system in executing the data transaction based on the component benchmark; assess a performance by the storage system of the data transaction to determine a performance metric; and compare the performance metric to the system benchmark to determine a remaining execution capacity of the storage system. In some embodiments, the machine executable code further causes the machine to provide an interface for displaying the remaining execution capacity of the storage system on a user display. In some embodiments, the machine executable code further causes the machine to move a data set from a first storage device to a second storage device based on the remaining execution capacity of the storage system.
The foregoing outlines features of several embodiments so that those skilled in the art may better understand the aspects of the present disclosure. Those skilled in the art should appreciate that they may readily use the present disclosure as a basis for designing or modifying other processes and structures for carrying out the same purposes and/or achieving the same advantages of the embodiments introduced herein. Those skilled in the art should also realize that such equivalent constructions do not depart from the spirit and scope of the present disclosure, and that they may make various changes, substitutions, and alterations herein without departing from the spirit and scope of the present disclosure.
Number | Name | Date | Kind |
---|---|---|---|
7599293 | Bain | Oct 2009 | B1 |
20080082844 | Ghiasi | Apr 2008 | A1 |
20090106008 | Branson | Apr 2009 | A1 |
20130312005 | Chiu | Nov 2013 | A1 |
20150201017 | Hrischuk | Jul 2015 | A1 |
Number | Date | Country | |
---|---|---|---|
20160134493 A1 | May 2016 | US |