). In any case, regardless of the measuring program, the IOPS parameter published by the manufacturer in the accompanying documentation does not guarantee the same performance in real conditions.
IOPS is measured by programs such as Iometer, originally developed by Intel, as well as IOzone, FIO and CrystalDiskMark.
Mainly, this indicator helps determine the device settings at which it displays maximum performance. The exact value of the parameter can vary greatly from system to system depending on the benchmark running conditions, including the ratio of reads and writes, the set of read blocks for sequential and random access, the number of threads and sampling depth, as well as the block size itself. There are other factors that affect the IOPS measurement result including system settings, device driver, applications running in background etc. In addition, before performing the test, you must read the manufacturer’s recommendations for the correct conduct of the test.
Performance Specifications
Sequential and random access to storage media
The main measured quantities are linear (sequential) and random (random) access operations. Linear read/write operations, in which parts of files are read sequentially, one after another, mean the transfer of large files (more than 128 K). During random operations, data is read randomly from different areas media, they are usually associated with a block size of 4 KB.
Below are the main characteristics:
For hard drives and other electromechanical storage devices, random access IOPS depends primarily on the device seek time, while in SSDs and storage systems based on them, the number of IOPS depends primarily on the operation of the internal microcontroller and the speed of the memory interface. On both device types, the number of IOPS in line operations (at a large block size) indicates the maximum throughput that can be achieved on the device. It is usually easier to show linear IOPS in MB/s:
IOPS *Block_size_in_bytes = Bytes_per_second(usually converted to MB/s)
While traditional hard disks have approximately the same number of IOPS for writing and reading, most SSDs based on NAND flash modules have significantly lower IOPS for writing than for reading due to the inability to write directly to the cell. Before this, you must perform a cleaning procedure (the so-called Garbage Collection).
Examples
Approximate IOPS values for hard drives:
| Device | Type | IOPS | Interface |
|---|---|---|---|
| 7,200 rpm SATA drives | HDD | ~75-100 IOPS | SATA 3 Gb/s |
| 10,000 rpm SATA drives | HDD | ~125-150 IOPS | SATA 3 Gb/s |
| 10,000 rpm SAS drives | HDD | ~140 IOPS | SAS |
| 15,000 rpm SAS drives | HDD | ~175-210 IOPS | SAS |
Approximate IOPS values for SSD
See also
Links
Performance measurement software
- Iometer Project - Official website of the Iometer project.
- CrystalDiskMark - Website of the CrystalDiskMark project.
Articles
- What is IOPS and what does it show - Article pc-hard.ru, 2011.
- IOPS fast calculation of storage systems for virtual infrastructure - Article, 2010.
- About performance: IOPS vs. MB/s - Article blog.aboutnetapp.ru, 2007.
Wikimedia Foundation. 2010.
See what "IOPS" is in other dictionaries:
IOPS- (del inglés Input/Output Operations Per Second, Instrucciones de Entrada/Salida Por Segundo), es una unidad de benchmark utilizada para medir el rendimiento de discos duros, unidades de estado sólido (SSD), memorias RAM y otras formas de… … Wikipedia Español
RAM drive, RAM disk (disk in memory), electronic disk computer technology, allowing you to store data in high-speed random access memory as on a block device (disk). Can be implemented both in software and in hardware.... ... Wikipedia
I/O-Performance
Input/Output-Performance- Input/Output operations Per Second, kurz IOPS, ist eine Benchmark Angabe von elektronischen Datenträgern. Sie gibt an wieviel Ein und Ausgabe Befehle pro Sekunde durchgeführt werden können. Dabei kann die Angabe genauer spezifiziert werden: Total ... Deutsch Wikipedia
HyperX is a division of Kingston Technology that specializes in producing high-performance products aimed at gamers and enthusiasts. It produces fast RAM modules, solid state drives, USB flash drives and computer headsets. Today I received one of the products from this series for testing - SSD HyperX FURY 240 GB.
Form factor:
2.5 inches
Interface:
SATA 3.0 (6Gbit/s), backward compatibility with SATA 2.0
Capacity:
240GB
Controller:
SandForce® SF-2281
Basic performance:
Compressible data transfer (ATTO) - 500 MB/s (read) and 500 MB/s (write)
Incompressible data transfer (AS-SSD and CrystalDiskMark):
470MB/s (read) and 220MB/s (write)
Maximum read/write speed of 4KB random blocks:
up to 84000/up to 41000 IOPS
Read/write speed of 4KB random blocks:
up to 22000/up to 41000 IOPS
PCMARK® Vantage HDD Suite Rating:
60000
PCMARK® 8 Storage Bandwidth Score:
180 MB/s
Total Bytes Written (TBW):
641 TB 2.5 DWPD
Energy consumption:
0.31 W Idle / 0.35 W Average / 1.65 W (Max) Read / 2.76 W (Max) Write
Storage temperature:
-40°C to 85°C
Working temperature:
from 0°C to 70°C
Size:
69.8 x 100.1 x 7 mm
Vibration during operation:
2.17 G (peak) at 7-800 Hz
Vibration when idle:
20 G (peak) at 10-2000 Hz
Expected service life:
1,000,000 hours (mean time between failures)
Warranty/Support:
three-year warranty and free technical support
Contents of delivery. Appearance.
Like most Kingston products, the hero of our today's review comes in cardboard and plastic packaging, the unauthorized opening of which will be immediately noticeable. Therefore, a potential buyer can be 100% sure that he is buying absolutely new product. The combination of red and black colors fully corresponds to the name of the drive - Fury (English rage), and makes you perceive the contents of the container as something fast and powerful.
The front of the package provides basic information about the SSD. The following indicators are indicated: capacity, maximum read and write speed, 3-year warranty information, as well as the connection interface - SATA 3.0. In addition, it is said that this instance is 16 times faster than usual hard drive with a spindle speed of 7200 rpm.
The reverse side also contains information about the drive.
In particular, it is said that the SSD increases the loading speed of applications and games installed on it. It is indicated that the drive is intended for use in work and home PCs and laptops, but is not suitable for use in server stations (due to its low resource). The actual capacity available to the user is less than declared. For those wondering why, Kingston directs them to their website. Unfortunately, the link only provides information on English language. The reason for this “incident” is quite clearly stated in Russian on Wikipedia.
Acquainted with complete information Under the Kingston warranty, you can again visit the official website.
In addition, at the bottom it is indicated that the drive uses an LSI SandForce controller and a sticker with markings is placed.
Upon opening the package we can find some more information about the warranty in several languages. Russian is also present.
The package itself is quite ascetic. Inside the plastic container there is the SSD itself, a frame with double-sided tape for the possibility of installing the drive in devices using 9.5 mm thick hard drives, as well as a sticker. The manufacturer did not consider it necessary to pamper the buyer with an adapter for 3.5", screws for fastening and other goodies. But this is understandable, top-end configurations are not inherent in budget solutions.
The case is made entirely of metal, which contributes to more efficient heat dissipation and adds strength to the structure.
The top of the drive is almost completely covered with a sticker, which not only provides basic information about the device, but also doubles as a warranty sticker, the removal of which will void the warranty.
Under the sticker there are four screws, which can only be unscrewed with a screwdriver with a Torx T-6H bit. I didn’t have it, so I didn’t bother disassembling it. If anyone is interested, then under the cover you will find 16 memory chips of 16 GB each and an LSI SandForce SF-2281 controller. The user has access to 240 GB instead of 256 due to the fact that 16 GB are allocated for reserve area, which is needed in order to extend the life of the drive.
There is nothing remarkable on the reverse side. There are holes for mounting.
The power and data connectors are quite standard.
Well, the thickness of the drive of 7 mm allows you to expand the scope of its application.
Test configuration:
Processor...................................Intel Core i5-3570
Motherboard...................Gigabyte GA-Z77-DS3H
RAM...................................Corsair Dominator Platinum CMD16GX3M4A2666C11 @ 2200MHz
SSD........................................Corsair Force GS 128 GB
Power supply........................Corsair AX760i
Operating system...............Windows 7 x64
Testing. General impressions.
After formatting, 223 GB becomes available to the user.
Let's take a look at the characteristics using the CrystalDiskInfo utility.
Already from the factory, our copy has the latest firmware.
SandForce-based drives can compress data before writing. On the one hand, this allows you to extend the life of the SSD by reducing the number of records, but on the other hand, we can observe a decrease in performance when working with poorly compressed data.
Let's check this statement in practice using the CrystalDiskMark program.
The left column will contain tests with default settings, and the right column will contain tests with well-compressed data.
So, the first three lines in each test window are most interesting to us. The top line (sequential reading/writing) reflects the work during normal file copying (music, video, archives, etc.). But the second and third lines (random reading/writing of 512 and 4 KB blocks) simulate the work operating system.
As we can see, the sequential reading speed when working with small files is significantly lower than with large ones. At the same time, when randomly reading/writing 4 KB blocks, performance, on the contrary, decreases as the size of the test file increases.
The recording speed when working with incompressible data dropped by 2 times, which was to be expected, and when working with compressible data, the speed indicators are close to the maximum according to the specification. The manufacturer described these nuances in technical specifications on the official website, but on the packaging only the maximum speed is indicated without indicating the testing conditions.
Compared to normal hard drive, then the performance of the SSD looks very impressive, without a doubt, any user will be able to notice the difference in the performance of the PC.
Let's continue testing using the AS SSD Benchmark utility. The data is not compressed here, so performance indicators are below the maximum, as in the previous case.
Additionally, you can see how quickly the drive will work with games, programs and copying large files (ISO image).
The speeds are not prohibitive, but quite acceptable.
Let's supplement the testing with AIDA64 Extreme readings.
In linear reading, there is one non-critical failure in the graphics:
Checking the access time showed excellent results:
Reading from the buffer is no problem:
A random reading, with the exception of a couple of dips in the graphics, gave good performance indicators:
Well, let’s finish checking our SSD using a program in which, according to the manufacturer, the drive shows maximum results:
And here everything completely falls into place. Well-compressed data = high results (even higher than Kingston claims).
The temperature during testing did not rise above 33 degrees Celsius, and during idle time it was even around 27 degrees. This is a big plus, because... The SSD does not heat up the internal space of the PC.
Conclusions.
Kingston took the path of least resistance by installing SandForce controllers in its drives due to attractive licensing conditions. SandForce works great with compressible data, but fails when recording files that are difficult to compress. At the same time, the speed indicators remain at the level of SSD drives on a similar controller from other manufacturers. The sufficiently large volume of the tested drive allows it to last longer than less capacious ones due to the reduced load on the memory cells. The manufacturer provides a 3-year warranty, but with typical use and 10+ years the drive should work without problems. For example, over the last six months I have only had 700 GB written to my system SSD.
Buy SSD drive as an alternative to a regular HDD or in addition to it? Undoubtedly yes. Should I buy HyperX FURY? It all depends on the price. Price tags may vary significantly at different retail outlets. Overall, this is a high-quality product with fairly high speed performance. Most inexperienced users will be pleasantly surprised by the responsiveness of the PC after using conventional HDDs. For example, loading the operating system takes about 14-17 seconds (depending on the number of programs in startup), launching any programs and games also happens in a matter of seconds. Well, what else does a person need to be happy?
According to our observations, assessing the performance of a storage system in IOps does not give an accurate idea of the performance of this storage system under load in 1C tasks, and we strongly recommend not using IOps to calculate equipment.
On the contrary, such an assessment may form a false impression of the high performance of the disk subsystem, while when used in productive conditions, it may become clear that the storage system is insufficiently powerful.
For example, can you tell what exactly an “operation” does from the IOps metric? How big is the data being read/written? Is it correct to compare IOps for regular hard drives and SSDs? How do methods for calculating IOps take into account the slowdown in speed on an SSD as it wears out or when there is little on the SSD? free space? Can you tell me the difference between Raw IOPS and Functional IOPS?
Note. Total Raw IOPS = Disk Speed IOPS * Number of disks
Functional IOPS =(((Total Raw IOPS×Write %))/(RAID Penalty))+(Total Raw IOPS×Read%)
Are you sure that different programs Will IOPS measurements give you the same results?
Programs for measuring IOPS
IOmeter - IOPS test
IOzone - IOPS test
FIO - IOPS test
CrystalDiskMark - IOPS test
SQLIO - a set of tests for calculating performance (IOPS, MB, Latency) for database servers
wmarow - RAID group calculator based on IOPS performance
Or let’s also say whether the IOps calculation method accurately takes into account disk response time and throughput?
To understand why not everything is simple, you need to consider a simple example and analogy.
Along the road, a large number of people must be moved from A to B. There are two options: we can transport them in their personal cars or put them on buses. The road capacity will, of course, be higher if people are transported by buses, that is, in “large blocks”. However, public transport methods usually conflict with individual goals and routes. It’s good if we have a huge plant in B, to which the main flow from A is directed. We can load all the “bytes” into one large package-bus at the entrance, and unload it at the stop at the plant, where all our “bytes” are actually sent.
However, if our bytes do not go to the factory, but travel on individual and independent business - “operations”, each with an individual route, then delivering them by “bus” - a large package will, on the contrary, lead to large losses of time. In this case, transportation by individual vehicles will be more profitable. However, the total capacity of a road filled with individual “packages” - cars carrying several bytes each, will, of course, be lower than when transported by a large package - “bus”.
Thus increase bandwidth in MB/s due to the enlargement of packets leads to a decrease in IOPS, and vice versa
, the increase in operations per second of “passengers delivered to the target” of our interface road, crowded with cars, leads to a decrease in its throughput in MB/s. You cannot simultaneously achieve high levels of IOPS and MB/s simply by physical properties existing equipment.
Either large “bus” packages and there are few of them (“operations per second”), or small individual “car” packages, each carrying out an individual “operation” to deliver data, but filling the entire road, and the overall human traffic as a result is small.
The choice of the necessary metrics is influenced by the nature of access to the data. Linear non-threaded disk access cannot be compared with highly concurrent and non-uniform random disk access.
To evaluate performance, we use observations of the current system and the level of equipment load, as well as queues on it during load peaks.
IOPS(number of input/output operations - from the English Input/Output Operations Per Second) is one of the key parameters when measuring the performance of data storage systems, hard drives (HDDs), solid-state drives (SSDs) and network storage (SAN).
In fact, IOPS is the number of blocks that can be counted or written to the media. The larger the block size, the fewer chunks that make up the file, and the lower the IOPS will be, since reading a larger chunk will take more time.
This means that to determine IOPS you need to know the speed and block size of a read/write operation. The IOPS parameter is equal to the speed divided by the block size of the operation.
Performance Specifications
The main measured quantities are linear (sequential) and random (random) access operations.
Linear read/write operations, in which parts of files are read sequentially, one after another, mean the transfer of large files (more than 128 K). Random operations read data randomly from different areas of the media and are typically associated with a 4 KB block size.
Below are the main characteristics:
Approximate IOPS values
Approximate IOPS values for hard drives.| Device |
Type |
IOPS |
Interface |
| 7,200 rpm SATA drives |
HDD |
~75-100 IOPS |
SATA 3 Gb/s |
| 10,000 rpm SATA drives |
HDD |
~125-150 IOPS |
SATA 3 Gb/s |
| 10,000 rpm SAS drives |
HDD |
~140 IOPS |
SAS |
| 15,000 rpm SAS drives |
HDD |
~175-210 IOPS |
SAS |
Approximate IOPS values for SSDs.
RAID penalty
Any read operations that are performed on the disks are not subject to any penalty since all disks can be used for read operations. But everything is reversed with write operations. The number of write penalties depends on the type of RAID selected, for example.In RAID 1, for data to be written to a disk, two write operations occur (one write to each disk), and therefore RAID 1 has two penalties.
In RAID 5, to write data, 4 operations occur (Read existing data, RAID parity, Write new data, Write new parity), thereby the penalty in RAID 5 is 4.
This table shows the penalty value for more commonly used RAID configurations.
Workload characteristics
Workload performance is generally considered as the percentage of reads and writes that an application generates or requires. For example, in a VDI environment, the IOPS percentage is considered to be 80-90% write and 10-20% read. Understanding the workload characteristics is the most critical factor because it will determine the optimal RAID for your environment. Write-intensive applications are good candidates for RAID 10, while read-intensive applications can be placed on RAID 5.IOPS calculation
There are two scenarios for calculating IOPS.One scenario is when we have a certain number of disks, and we want to know how many IOPS these disks will produce?
The second scenario, when we know how many IOPS we need, and want to calculate the required number of disks?
Scenario 1: Calculating IOPS based on a certain number of disks
Let's imagine that we have 20 450GB 15k RPM disks. Consider two workload scenarios of 80%Write-20%Read and another scenario with 20%Write-80%Read. We will also calculate the number of IOPS for both RAID5 and RAID 10.Formula for calculating IOPS:
Total Raw IOPS = Disk Speed IOPS * Number of disks
Functional IOPS =(((Total Raw IOPS×Write %))/(RAID Penalty))+(Total Raw IOPS×Read%)
There is a definition of Raw IOPS and Functional IOPS; the Functional IOPS currents are those IOPS that include RAID penalties, and these are the “real” IOPS.
Now let’s plug in the numbers and see what happens.
Total Raw IOPS= 170*20 = 3400 IOPS (one 15K RPM disk can produce an average of 170 IOPS)
For RAID-5
Option 1 (80%Write 20%Read) Functional IOPS = (((3400*0.8))/(4))+(3400*0.2) = 1360 IOPSOption 2 (20%Write 80%Read) Functional IOPS = (((3400*0.2))/(4))+(3400*0.8) = 2890 IOPS
For RAID-1
Option 1 (80%Write 20%Read) Functional IOPS = (((3400*0.8))/(2))+(3400*0.2) = 2040 IOPSOption 2 (20%Write 80%Read) Functional IOPS = (((3400*0.2))/(2))+(3400*0.8) = 3100 IOPS
Scenario 2: Counting the number of disks to achieve a certain number of IOPS
Consider a situation where we need to determine the RAID type and the number of disks to achieve a certain number of IOPS 5000 and with certain workloads, for example 80%Write20%Read and 20%Write80% Read.Again, to begin with, the formula by which we will calculate:
Total number of Disks required = ((Total Read IOPS + (Total Write IOPS*RAID Penalty))/Disk Speed IOPS)
Total IOPS = 5000
Now let's plug in the numbers.
The note: 80% of 5000 IOPS = 4000 IOPS and 20% of 5000 IOPS = 1000 IOPS we will operate with these numbers.
For RAID-5
Option 1 (80%Write20%Read) – Total Number of disks required = ((1000+(4000*4))/170) = 100 disks.Option 2 (20%Write80%Read) – Total Number of disks required = ((4000+(1000*4))/170) = 47 disks approximately.
For RAID-1
Option 1 (80%Write20%Read) – Total Number of disks required = ((1000+(4000*2))/170) = 53 disks approximately.Option 2 (20%Write80%Read) – Total Number of disks required = ((4000+(1000*2))/170) = 35 disks approximately.
Understanding and calculating IOPS, RAID penalties, and workload characteristics is a very critical aspect of planning. When the write load is more intense, choose RAID 10 and vice versa for read loads, RAID 5.
IOPS used to determine the performance of a disk or disk array.
IOPS means Input/Output (operations) Per Second , quantity “I/O operations per second”. The quantity measures the amount of work done over a certain period of time. In fact, IOPS This is the number of blocks that can be counted or written to the media. The larger the block size, the fewer pieces the file consists of, and the smaller the file will be IOPS, since reading a larger piece will take more time.
“I/O operation”- this is simply a certain part of the work of the disk subsystem, which is performed in response to a request from the host server and/or some internal processes. This is usually read or write with various subcategories e.g. "reading" (read), “re-reading” (re-read), “record”(write), “overwrite” ( re-write), “random access type” (random), “serial access type” (sequential) and the size of the data block being operated.
The main measured quantities are linear (sequential) and random (random) access operations.
Linear read/write operations, in which parts of files are read sequentially, one after another, mean the transfer of large files (more than 128 K). Random operations read data randomly from different areas of the media and are typically associated with a 4 KB block size.
Depending on the type of operation, this size can vary from bytes to kilobytes and even several megabytes. There are many types of I/O, and a multitasking and multihost system almost never uses just one. Virtualization only adds variety to I/O patterns.
No storage system can show maximum values IOPS regardless of the nature of the input/output operations, values latency and block size.
Latency it is a measure of how long a single I/O request takes to complete, from the application's perspective.
Significant volumes I/O wait this is a sign that the source of the problem is storage (there are other sources of delays, CPU and network are common examples). Even in case of good performance latency if you see a large number I/O waits- this means that the application would like more speed from the storage system.
Determining disk system performance is an often overlooked aspect of system design. Since the disk system is the slowest medium on a computer, it should be one of the FIRST components to be specified correctly.
Write-intensive applications are good candidates for RAID 10, while applications that intensively use read operations can be placed on RAID 5.
IOPS used to determine the performance of a disk or disk array. For example, we can assume that the maximum IOPS for disk:
To calculate IOPS we use the equation:
| IOPS = 1/(avgLatency + avgSeek) IOPS = 1/(0.00416 + 0.0085) = 78.9889415 |
Total, maximum IOPS - 79.
Calculate the maximum IOPS value for a disk array
As a note to storage system design, calculating the performance of a disk system is critical to the performance of a given system. Most systems use RAID to provide storage redundancy. This section describes how to calculate IOPS For RAID-arrays.
Maximum Read IOPS
IOPS reading ( maxReadIops) For RAID-array:
maxReadIops = numDisks * diskMaxIops
Accordingly, for an array of 4 disks maximum value IOPS reading will be as follows:
| maxReadIops = 4 * 79 maxReadIops = 316 |
Maximum Write IOPS
Calculating the maximum value IOPS records ( maxWriteIops) - this is completely different in relation RAID-arrays. RAID-arrays have a write penalty, and the type RAID-array determines the severity of the penalty. This penalty is a result of the redundancy provided RAID, since the array must necessarily write data to multiple disks/locations to ensure data integrity.
Penalty for writing a RAID array
Most common types RAID and their recording penalties are determined in the following table:
Total, for our example, the maximum value IOPS to write for an array RAID 10 - 158.
Design for Performance
Simple calculation maximum quantity IOPS to read and write for existing or future RAID-array is not enough. To ensure consistent and sustainable performance, it is necessary to determine the performance requirements for the system to determine The best decision for disk. Minimum Required IOPS must be defined in such a way that the required number of disks can be purchased at the required speed.
First you need to know the performance requirements (for example, read and write IOPS) for a given system or application. This information can be obtained from the supplier's documentation or software.
Calculating the minimum required IOPS
Let's say we have an application that requires 600 Read IOPS and
300 Write IOPS. The disk array is assembled in RAID 5.
To calculate the minimum quantity IOPS (minReqdIops), add the quantity required IOPS reading ( reqdReadIops) to the sum of the quantities required IOPS records ( reqdWriteIops) and fine RAID (raidWritePenalty): minReqdIops = reqdReadIops + (reqdWriteIops * raidWritePenalty)
In our example:
| minReqdIops = 600 + (300 * 4) minReqdIops = 1800 |
Minimal amount IOPS, required to ensure the performance level for our example is 1800.
NOTE. This calculation determines the minimum quantity IOPS required to meet the performance specification. This means that the disk array should NOT operate below this performance level.
Calculating the minimum number of disks for a RAID array
Once the minimum quantity required IOPS defined, it is very easy to determine the minimum number and speed of disks required to create RAID-array to meet performance requirements.
Minimum number of disks by disk speed
The minimum number of disks required to meet our performance requirement ( minNumDiskMinPerf), calculated as follows: minNumDisksMinPerf = minReqdIops / maxIopsByDiskSpeed
Using information based on the minimum required IOPS above and assuming we want to create an array of 10,000 RPM-disks ( ~125-150 IOPS), calculating the minimum number of disks that will match our minimum requirements to productivity ( minNumDisksMinPerf) 1800 IOPS (minReqdIops) as follows:
| minNumDisksMinPerf = 1800 / 130 minNumDisksMinPerf = 14 |
Minimum number of disks 10,000 RPM required to meet our performance requirements is 14.
Minimum number of disks by RAID type
Type RAID defines the minimum number of disks to satisfy the type requirements RAID. For example, for RAID 5 always requires at least 3 disks. For RAID 10 always requires at least 4 disks.
For any arrays requiring large quantity drives, use the multiplier in the table below to determine the correct number of drives to meet the type requirements RAID:
After calculating the number of disks by speed, we determine the minimum number of disks required by type RAID.
In the example where 10K RPM drives were chosen to build the array, the calculation shows that at least 14 drives are required. If type RAID there will be 5, 14 disks will be sufficient. However, if the type RAID will be 10, the minimum number of disks required by this type RAID, will be 8, since the multiplier for RAID 10 is equal to 4.
Programs for measuring IOPS
IOmeter - test IOPS
IOzone - test IOPS
FIO - test IOPS
CrystalDiskMark - test IOPS
SQLIO - a set of tests for calculating performance ( IOPS, M.B., Latency) for database servers
wmarow - calculator RAID in terms of performance IOPS