View source for IOPS

{{short description|Input/output performance measurement}}
'''Input/output operations per second''' ('''IOPS''', pronounced ''eye-ops'') is an [[input/output]] performance measurement used to characterize [[data storage device|computer storage]] devices like [[hard disk drive]]s (HDD), [[solid state drives]] (SSD), and [[storage area network]]s (SAN). Like [[Benchmark (computing)|benchmark]]s, IOPS numbers published by storage device manufacturers do not directly relate to real-world application performance.<ref name=Scott_Lowe>{{cite web |url=http://www.techrepublic.com/blog/datacenter/calculate-iops-in-a-storage-array/2182 |title=Calculate IOPS in a storage array |last=Lowe |first=Scott |publisher=techrepublic.com |date=2010-02-12 |access-date=2011-07-03}}</ref><ref name="Symantec">{{cite web |url=https://community.broadcom.com/symantecenterprise/communities/community-home/librarydocuments/viewdocument?DocumentKey=e6fb4a1b-fa13-4956-b763-8134185c0c0a&CommunityKey=63b01f30-d5eb-43c7-9232-72362b508207&tab=librarydocuments |title=Getting The Hang Of IOPS v1.3 |date=2012-08-03 |access-date=2013-08-15}}</ref>

==Background==
To meaningfully describe the performance characteristics of any storage device, it is necessary to specify a minimum of three metrics simultaneously: IOPS, response time, and (application) workload. Absent simultaneous specifications of response-time and workload, IOPS are essentially meaningless. In isolation, IOPS can be considered analogous to "revolutions per minute" of an automobile engine i.e. an engine capable of spinning at 10,000 RPMs with its transmission in neutral does not convey anything of value, however an engine capable of developing specified torque and horsepower at a given number of RPMs fully describes the capabilities of the engine.

The specific number of IOPS possible in any system configuration will vary greatly, depending upon the variables the tester enters into the program, including the balance of read and write operations, the mix of [[Sequential access|sequential]] and [[Random access|random]] access patterns, the number of worker [[Thread (computer science)|threads]] and queue depth, as well as the data block sizes.<ref name=Scott_Lowe/> There are other factors which can also affect the IOPS results including the system setup, storage drivers, OS background operations etc. Also, when testing SSDs in particular, there are preconditioning considerations that must be taken into account.<ref name=Kent_Smith>{{cite web |url=http://www.flashmemorysummit.com/English/Collaterals/Proceedings/2009/20090811_F2A_Smith.pdf |title=Benchmarking SSDs: The Devil is in the Preconditioning Details |last= Smith |first=Kent |publisher=SandForce.com |date=2009-08-11 |access-date=2015-05-05}}</ref>

==Performance characteristics==
[[Image:Random vs sequential access.svg|thumb|right|Random access compared to sequential access.]]The most common performance characteristics measured are sequential and random operations. Sequential operations access locations on the storage device in a [[Contiguous data storage|contiguous manner]] and are generally associated with large data transfer sizes, e.g. ≥ 128&nbsp;[[Kilobyte|kB]]. Random operations access locations on the storage device in a non-contiguous manner and are generally associated with small data transfer sizes, e.g. 4&nbsp;kB.

The most common performance characteristics are as follows:

{| class="wikitable"
|-
! scope="col" | Measurement
! scope="col" | Description
|-
! scope="row" | Total IOPS
| Total number of I/O operations per second (when performing a mix of read and write tests)
|-
! scope="row" | Random Read IOPS
| Average number of random read I/O operations per second
|-
! scope="row" | Random Write IOPS
| Average number of random write I/O operations per second
|-
! scope="row" | Sequential Read IOPS
| Average number of sequential read I/O operations per second
|-
! scope="row" | Sequential Write IOPS
| Average number of sequential write I/O operations per second
|}

For HDDs and similar electromechanical storage devices, the random IOPS numbers are primarily dependent upon the storage device's random [[seek time]], whereas, for SSDs and similar solid state storage devices, the random IOPS numbers are primarily dependent upon the storage device's internal controller and memory interface speeds. On both types of storage devices, the sequential IOPS numbers (especially when using a large block size) typically indicate the maximum sustained bandwidth that the storage device can handle.<ref name=Scott_Lowe/> Often sequential IOPS are reported as a simple [[Data-rate units#Megabyte per second|Megabytes per second]] number as follows:

<blockquote><math>\text{IOPS} \times \text{TransferSizeInBytes} = \text{BytesPerSec}</math> (then converted to [[Data-rate units#Megabyte per second|MB/s]])
</blockquote>

Some HDDs/SSDs will improve in performance as the number of outstanding IOs (i.e. queue depth) increases. This is usually the result of more advanced controller logic on the drive performing command queuing and reordering commonly called either [[Tagged Command Queuing]] (TCQ) or [[Native Command Queuing]] (NCQ). Many consumer [[SATA]] HDDs either cannot do this, or their implementation is so poor that no performance benefit can be seen.{{Citation needed|date=August 2008}} Enterprise class SATA HDDs, such as the [[Western Digital Raptor]] and Seagate Barracuda NL will improve by nearly 100% with deep queues.<ref>{{cite web |url=http://www.storagereview.com/articles/200607/500_6.html |title=SATA in the Enterprise - A 500 GB Drive Roundup &#124; StorageReview.com - Storage Reviews |publisher=StorageReview.com |date=2006-07-13 |access-date=2013-05-13 |archive-date=2013-05-05 |archive-url=https://web.archive.org/web/20130505050424/http://www.storagereview.com/articles/200607/500_6.html |url-status=dead }}</ref> High-end [[SCSI]] drives more commonly found in servers, generally show much greater improvement, with the [[Seagate Technology|Seagate]] Savvio exceeding 400 IOPS—more than doubling its performance.{{Citation needed|date=August 2008}}

While traditional HDDs have about the same IOPS for read and write operations, many [[Flash memory|NAND flash-based]] SSDs and USB sticks are much slower writing than reading due to the inability to rewrite directly into a previously written location forcing a procedure called [[Garbage collection (SSD)|garbage collection]].<ref name="IBM_WA">{{cite journal |title=Write Amplification Analysis in Flash-Based Solid State Drives |author1=Xiao-yu Hu |first2=Evangelos |last2=Eleftheriou |first3=Robert |last3=Haas |first4=Ilias |last4=Iliadis |first5=Roman |last5=Pletka |year=2009 |publisher=[[IBM]] | citeseerx = 10.1.1.154.8668 }}</ref><ref name="OCZ_WA">{{cite web |url=http://www.oczenterprise.com/whitepapers/ssds-write-amplification-trim-and-gc.pdf |title=SSDs - Write Amplification, TRIM and GC |publisher=OCZ Technology |access-date=2010-05-31 |archive-url=https://web.archive.org/web/20120526033947/http://www.oczenterprise.com/whitepapers/ssds-write-amplification-trim-and-gc.pdf |archive-date=2012-05-26 |url-status=dead }}</ref><ref>{{cite web |url=http://www.intel.com/cd/channel/reseller/asmo-na/eng/products/nand/feature/index.htm |title=Intel Solid State Drives |publisher=Intel |access-date=2010-05-31}}</ref> This has caused hardware test sites to start to provide independently measured results when testing IOPS performance.

Flash SSDs, such as the Intel X25-E (released 2010), have much higher IOPS than traditional HDD. In a test done by Xssist, using [[Iometer]], 4&nbsp;KB random transfers, 70/30 read/write ratio, queue depth 4, the IOPS delivered by the Intel X25-E 64&nbsp;GB G1 started around 10000 IOPs, and dropped sharply after 8 minutes to 4000 IOPS, and continued to decrease gradually for the next 42 minutes. IOPS vary between 3000 and 4000 from approximately 50 minutes and onwards, for the rest of the 8+ hours the test ran.<ref>{{cite web |url=http://www.xssist.com/blog/Intel%20X25-E%2064GB%20G1,%204KB%2070%2030%20RW%20Random%20IOPS,%20iometer%20benchmark.htm |title=Intel X25-E 64GB G1, 4KB Random IOPS, iometer benchmark |date=2010-03-27 |access-date=2010-04-01}}</ref> Even with the drop in random IOPS after the 50th minute, the X25-E still has much higher IOPS compared to traditional hard disk drives. Some SSDs, including the [[OCZ]] RevoDrive 3 x2 PCIe using the [[SandForce]] controller, have shown much higher sustained write performance that more closely matches the read speed.<ref>{{cite web |url=http://thessdreview.com/our-reviews/ocz-revodrive-3-x2-480-gb-pcie-ssd-review-1-5gb-read1-25gb-write200000-iops-for-699/ |title=OCZ RevoDrive 3 x2 PCIe SSD Review – 1.5GB Read/1.25GB Write/200,000 IOPS As Little As $699 |date=2011-06-28 |access-date=2011-06-30}}</ref>  For example, a typical operating system has many small files (such as DLLs ≤ 128 kB), so SSD is more suitable for system drive.<ref>{{Cite web |title=SSD vs. HDD: Which Do You Need? |url=https://www.avast.com/c-ssd-vs-hdd |access-date=2024-06-24 |website=SSD vs. HDD: Which Do You Need? |language=en}}</ref>

==Examples==

=== Mechanical hard drives ===
Block size used when testing significantly affects the number of IOPS performed by a given drive. See below for some typical performance figures:<ref>{{Cite web|url=https://wintelguy.com/raidperf.pl|title=RAID Performance Calculator - WintelGuy.com|website=wintelguy.com|access-date=2019-04-01}}</ref>

{| class="wikitable sortable"
|-
! scope="col" |Drive (Type / RPM)
! scope="col" |IOPS 
(4&nbsp;KB block, random)
! scope="col" |IOPS 
(64&nbsp;KB block, random)
! scope="col" |MB/s (64KB block, random)
! scope="col" |IOPS 
(512&nbsp;KB block, random)
!MB/s (512&nbsp;KB block, random)
!MB/s (large block, sequential)
|-
|[[Serial Attached SCSI|SAS]] / 15K
|188 - 203
|175 - 192
|11.2 – 12.3
|115 – 135
|58.9 – 68.9
|91.5 – 126.3
|-
|[[Fibre Channel|FC]] / 15K
|163 - 178
|151 - 169
|9.7 – 10.8
|97 – 123
|49.7 – 63.1
|73.5 – 127.5
|-
|[[Fibre Channel|FC]] / 10K
|142 - 151
|130 – 143
|8.3 – 9.2
|80 – 104
|40.9 – 53.1
|58.1 – 107.2
|-
|[[Serial Attached SCSI|SAS]] / 10K
|142 - 151
|130 – 143
|8.3 – 9.2
|80 – 104
|40.9 – 53.1
|58.1 – 107.2
|-
|[[Serial ATA|SATA]] / 7200
|73 - 79
|69 - 76
|4.4 – 4.9
|47 – 63
|24.3 – 32.1
|43.4 – 97.8
|-
|[[Serial ATA|SATA]] / 5400
|57
|55
|3.5
|44
|22.6
|<!-- | 10,000 rpm [[SATA]] drives, queue depth 24
| HDD
| ~290 IOPS
| SATA 3 Gb/s
| <code>fio -readonly -name iops -rw=randread -bs=512 -runtime=20 -iodepth 24 -filename /dev/sda -ioengine libaio -direct=1</code>
|-  -->
|}

=== Solid-state devices ===
{{Example farm|section|date=May 2018}}
{| class="wikitable sortable"
|-
! scope="col" | Device
! scope="col" | Type
! scope="col" | IOPS
! scope="col" | Interface
! scope="col" | Notes
|-
| [[Intel#Solid-state drives (SSD)|Intel X25-M G2]] ([[Multi-level cell|MLC]])
| [[Solid-state drive|SSD]]
| ~8,600 IOPS<ref>{{cite web |url=http://www.tomshardware.com/reviews/Intel-x25-m-SSD,2012.html |title=Intel's X25-M Solid State Drive Reviewed                 |last1=Schmid |first1=Patrick |last2=Roos |first2=Achim |date=2008-09-08 |access-date=2011-08-02}}</ref>
| SATA 3 Gbit/s
| Intel's data sheet<ref>{{cite web |url=http://download.intel.com/design/flash/nand/mainstream/322296.pdf |title=Intel X18-M/X25-M SATA Solid State Drive — 34 nm Product Line |date=January 2010 |publisher=[[Intel]] |access-date=2010-07-20 |url-status=dead |archive-url=https://web.archive.org/web/20100812001346/http://download.intel.com/design/flash/nand/mainstream/322296.pdf |archive-date=2010-08-12 }}</ref> claims 6,600/8,600 IOPS (80&nbsp;GB/160&nbsp;GB version) and 35,000 IOPS for random 4&nbsp;KB writes and reads, respectively.
|-
| [[Intel#Solid-state drives (SSD)|Intel X25-E]] (SLC)
| SSD
| ~5,000 IOPS<ref>{{cite web| first1=Patrick|last1=Schmid |first2=Achim|last2=Roos |url=http://www.tomshardware.com/reviews/intel-x25-e-ssd,2158.html |title=Intel's X25-E SSD Walks All Over The Competition: They Did It Again: X25-E For Servers Takes Off |website=TomsHardware.com |date=27 February 2009 |access-date=2013-05-13}}</ref>
| SATA 3 Gbit/s
| Intel's data sheet<ref>{{cite web |url=http://download.intel.com/design/flash/nand/extreme/extreme-sata-ssd-datasheet.pdf |title=Archived copy |access-date=2009-03-18 |url-status=dead |archive-url=https://web.archive.org/web/20090206005330/http://download.intel.com/design/flash/nand/extreme/extreme-sata-ssd-datasheet.pdf |archive-date=2009-02-06 }}</ref> claims 3,300 IOPS and 35,000 IOPS for writes and reads, respectively. 5,000 IOPS are measured for a mix. Intel X25-E G1 has around 3 times higher IOPS compared to the Intel X25-M G2.<ref>{{cite web |url=http://www.xssist.com/blog/%5BSSD%5D_Comparison_of_Intel_X25-E_G1_vs_Intel_X25-M_G2.htm |title=Intel X25-E G1 vs Intel X25-M G2 Random 4&nbsp;KB IOPS, iometer |date=May 2010 |access-date=2010-05-19}}</ref>
|-
| [[G.Skill]] Phoenix Pro
| SSD
| ~20,000 IOPS<ref name="tweakpc">{{cite web|url=http://www.tweakpc.de/hardware/tests/ssd/gskill_phoenix_pro/s05.php |title=G.Skill Phoenix Pro 120 GB Test - SandForce SF-1200 SSD mit 50K IOPS - HD Tune Access Time IOPS (Diagramme) (5/12) |publisher=Tweakpc.de |access-date=2013-05-13}}</ref>
| SATA 3 Gbit/s
| [[SandForce]]-1200 based SSD drives with enhanced firmware, states up to 50,000 IOPS, but benchmarking shows for this particular drive ~25,000 IOPS for random read and ~15,000 IOPS for random write.<ref name="tweakpc" />
|-
| [[Samsung Electronics|Samsung]] SSD 850 PRO
| SSD
| Up to{{pov inline|date=January 2024}} 100,000 read IOPS<br />Up to{{pov inline|date=January 2024}} 90,000 write IOPS<ref name=Samsung-SSD850PRO>{{cite web|title=Samsung SSD 850 PRO Specifications|url=http://www.samsung.com/semiconductor/minisite/ssd/product/consumer/850pro.html|publisher=[[Samsung Electronics]]|access-date=7 June 2017}}</ref>{{primary source inline|date=January 2024}}
| SATA 6 Gbit/s
| 4&nbsp;KB aligned random I/O at QD32<br />10,000 read IOPS, 36,000 write IOPS at QD1<br />550&nbsp;MB/s sequential read, 520&nbsp;MB/s sequential write on 256&nbsp;GB and larger models<br />550&nbsp;MB/s sequential read, 470&nbsp;MB/s sequential write on 128&nbsp;GB model<ref name=Samsung-SSD850PRO />
|-
| [[Virident Systems]] tachIOn
| SSD
| Up to{{pov inline|date=January 2024}} 320,000 sustained READ IOPS using 4&nbsp;kB blocks, and up to 200,000 sustained WRITE IOPS using 4kB blocks<ref>{{cite news|url=https://www.theregister.co.uk/2010/06/16/virident_tachion/|title=Virident's tachIOn SSD flashes by|work=theregister.co.uk}}</ref>
| PCIe
|
|-
| OCZ RevoDrive 3 X2
| SSD
| Up to{{pov inline|date=January 2024}} 200,000 Random Write 4k IOPS<ref>{{cite web|url=http://www.storagereview.com/ocz_revodrive_3_x2_480gb_review |title=OCZ RevoDrive 3 X2 480GB Review &#124; StorageReview.com - Storage Reviews |publisher=StorageReview.com |date=2011-06-28 |access-date=2013-05-13}}</ref>
| PCIe
|
|-
| [[DDRdrive]] X1
| SSD
| Up to{{pov inline|date=January 2024}} 300,000+ (512B Random Read IOPS) and 200,000+ (512B Random Write IOPS)<ref>http://www.ddrdrive.com/ddrdrive_press.pdf {{Bare URL PDF|date=March 2022}}</ref><ref>{{Cite web |url=http://www.ddrdrive.com/ddrdrive_brief.pdf |title=Archived copy |access-date=2009-05-22 |archive-url=https://web.archive.org/web/20090520234015/http://www.ddrdrive.com/ddrdrive_brief.pdf |archive-date=2009-05-20 |url-status=dead }}</ref><ref>{{Cite web |url=http://www.ddrdrive.com/ddrdrive_bench.pdf |title=Archived copy |access-date=2009-05-22 |archive-url=https://web.archive.org/web/20090520235817/http://www.ddrdrive.com/ddrdrive_bench.pdf |archive-date=2009-05-20 |url-status=dead }}</ref>{{primary source inline|date=January 2024}}<ref>{{cite web |author=Allyn Malventano |url=http://www.pcper.com/reviews/Storage/DDRdrive-hits-ground-running-PCI-E-RAM-based-SSD |title=DDRdrive hits the ground running - PCI-E RAM-based SSD &#124; PC Perspective |publisher=Pcper.com |date=2009-05-04 |access-date=2013-05-13 |archive-url=https://web.archive.org/web/20130714005308/http://www.pcper.com/reviews/Storage/DDRdrive-hits-ground-running-PCI-E-RAM-based-SSD |archive-date=2013-07-14 |url-status=dead }}</ref>
| PCIe
|
|-
| [[Samsung Electronics|Samsung]] SSD 960 EVO
| SSD
| Up to{{pov inline|date=January 2024}} 380,000 read IOPS<br />Up to{{pov inline|date=January 2024}} 360,000 write IOPS<ref name="Samsung-SSD960">{{cite web|author1=Samsung Electronics|title=NVMe SSD 960 PRO/EVO|url=http://www.samsung.com/semiconductor/minisite/ssd/product/consumer/ssd960.html|access-date=7 June 2017|author1-link=Samsung Electronics}}</ref>{{primary source inline|date=January 2024}}
| [[NVM Express|NVMe]] over PCIe 3.0 x4, [[M.2]]
| 4 kB aligned random I/O with four workers at QD4 (effectively QD16),<ref name="TomsHardware-SamsungSSD960PRO">{{cite web|last1=Ramseyer|first1=Chris|title=Samsung 960 Pro SSD Review|url=http://www.tomshardware.com/reviews/samsung-960-pro-ssd-review,4774.html|website=[[Tom's Hardware]]|date=18 October 2016|access-date=9 June 2017|quote=Samsung tests NVMe products with four workers at QD4}}</ref> 1 TB model<br />14,000 read IOPS, 50,000 write IOPS at QD1<br />330,000 read IOPS, 330,000 write IOPS on 500 GB model<br />300,000 read IOPS, 330,000 write IOPS on 250 GB model<br />Up to{{pov inline|date=January 2024}} 3.2&nbsp;GB/s sequential read, 1.9&nbsp;GB/s sequential write<ref name="Samsung-SSD960" />
|-
| [[Samsung Electronics|Samsung]] SSD 960 PRO
| SSD
| Up to{{pov inline|date=January 2024}} 440,000 read IOPS<br />Up to{{pov inline|date=January 2024}} 360,000 write IOPS<ref name="Samsung-SSD960" />{{primary source inline|date=January 2024}}
| [[NVM Express|NVMe]] over PCIe 3.0 x4, [[M.2]]
| 4kB aligned random I/O with four workers at QD4 (effectively QD16),<ref name="TomsHardware-SamsungSSD960PRO" /> 1 TB and 2 TB models<br />14,000 read IOPS, 50,000 write IOPS at QD1<br />330,000 read IOPS, 330,000 write IOPS on 512 GB model<br />Up to{{pov inline|date=January 2024}} 3.5&nbsp;GB/s sequential read, 2.1&nbsp;GB/s sequential write<ref name="Samsung-SSD960" />
|-
| [[Kaminario]] K2
| SSD
| Up to{{pov inline|date=January 2024}} 2,000,000 IOPS.<ref name="Storage Review">{{cite web|url=http://www.storagereview.com/kaminario_boasts_over_2_million_iops_and_20_gb_s_throughput_on_a_single_allflash_k2_storage_system|title=Kaminario Boasts Over 2 Million IOPS and 20 GB/s Throughput on a Single All-Flash K2 Storage System|author=Lyle Smith|access-date=2015-12-19|archive-date=2016-04-22|archive-url=https://web.archive.org/web/20160422215020/http://www.storagereview.com/kaminario_boasts_over_2_million_iops_and_20_gb_s_throughput_on_a_single_allflash_k2_storage_system|url-status=dead}}</ref><br />1,200,000 IOPS in SPC-1 benchmark simulating business applications<ref name="The Register">{{cite web|last=Mellor |first=Chris |url=https://www.theregister.co.uk/2012/07/30/kaminario_spc_1/ |title=Chris Mellor, The Register, July 30, 2012: "Million-plus IOPS: Kaminario smashes IBM in DRAM decimation" |publisher=Theregister.co.uk |date=2012-07-30 |access-date=2013-11-14}}</ref><ref name="Storage Performance Council">{{cite web|url=http://www.storageperformance.org/results/benchmark_results_spc1/#kaminario_spc1|title=Storage Performance Council: Active SPC-1 Results|author=Storage Performance Council|work=storageperformance.org|access-date=2012-09-25|archive-url=https://web.archive.org/web/20140925233025/http://www.storageperformance.org/results/benchmark_results_spc1#kaminario_spc1|archive-date=2014-09-25|url-status=dead}}</ref>
| [[Fibre Channel|FC]]
| [[Multi-level cell|MLC Flash]]
|-
| [[NetApp]] FAS6240 cluster
| Flash/Disk
| 1,261,145 SPECsfs2008 nfsv3 IOPs using 1,440 15k disks, across 60 shelves, with virtual storage tiering.<ref>{{cite web|url=http://www.spec.org/sfs2008/results/res2011q4/sfs2008-20111017-00199.html |title=SpecSFS2008 |access-date=2014-02-07}}</ref>{{unreliable source|date=January 2024}}
| NFS, SMB, FC, FCoE, iSCSI
| SPECsfs2008 is the latest version of the Standard Performance Evaluation Corporation benchmark suite measuring file server throughput and response time, providing a standardized method for comparing performance across different vendor platforms.
|-
| [[EMC Corporation|EMC]] DSSD D5
| Flash
| Up to{{pov inline|date=January 2024}} 10 million IOPS<ref>{{cite web|url=https://www.emc.com/en-us/storage/flash/dssd/dssd-d5/index.htm |title=Rack-Scale Flash Appliance - DSSD D5 EMC |publisher=EMC |access-date=2016-03-23}}</ref>{{primary source inline|date=January 2024}}
| PCIe Out of Box, up to 48 clients with [[high availability]].
| PCIe Rack Scale Flash Appliance. Product discontinued.<ref>{{Cite news|url=https://www.theregister.co.uk/2017/03/02/dell_cans_standalone_dssd/|title = Dell kills off standalone DSSD D5, scatters remains into other gear| website=[[The Register]] }}</ref>
|}

==See also==
* [[Instructions per second]]
* [[Performance per watt]]

==References==
{{Reflist|30em}}

{{Solid-state Drive}}

[[Category:Benchmarks (computing)]]
[[Category:Data transmission]]
[[Category:Units of frequency]]