Iometer Storage Test Tool

IOMETER

Official website: http://www.iometer.org/

Iometer's user guide says that Iometer is a stand-alone or cluster I/O subsystem measurement and Description tool. It was initially announced by Intel's Intel Developer Forum (IDF) on February 17, 1998, and has since been widely spread across the industry as a standard. At the same time, Intel has stopped Iometer development and handed the open source to osdl. Iometer was registered in SourceForge.net in November 2001, and the development project was restarted from 2003 to February, and the project maintainer became an independent internal working group, of course. (Not Intel)

Compared with the previous 3 test software, Iometer belongs to the test software more free, can according to user needs to adjust the scope of the test, data block size, processing mode and so on, of course, beginners will be more dizzy at first. Iometer Current stable version of 2006, and SandForce control of the SSD vendors are generally like to use Iometer 2008 to do the test (I will speak later), the latest version for IOMETER-1.1.0-RC1. Here I'll take the latest version to explain. (More options)


Iometer includes 2 programs, Iometer.exe and Dynamo.exe. Which Iometer is a control program, is a graphical interface, so that you easily adjust the parameters and display test results, and Dynamo is to allow the test disk to generate pressure test of the main program, with Iometer to control the Dynamo program. When you run Iometer under Windows, you automatically open the Dynamo program (you'll notice that 2 windows are open, and the DOS command line program is Dynamo.) ）


1) Topology/workers:

The number of workers depends on the number of CPU cores on the current machine (my CPU is the 2600k,4 kernel 8 thread, so there are 8 workers), each worker can let you set the number of threads that the hard drive executes simultaneously, but generally we run a household test with only one worker. So you need only 1 worker, multithreaded testing for high-performance servers or enterprise-class storage devices. We can remove the extra worker here before the test.

2) Targets:(There are 3 different icons to indicate different types of hard drives when setting disk target)

The yellow icon represents "Logical Drive Logical Disk", meaning the disk that has been divided, formatted, and has a file system above it. (Displays the hard disk partition volume label and the volume label name)

If a red strikethrough is added to the yellow icon, it means that there is no iobw.tst file on the logical hard drive. Iometer before testing a logical hard disk, a IOBW.TST test file is generated that calculates IO scores and transfer rates by "read/write" operations on the file.

The blue icon indicates the original (raw) hard drive, meaning that the file system on the hard drive is unrecognized or not partitioned and formatted, and the test file cannot be created on it. (The model of the hard disk will be displayed)

One thing to note here is that if you've done a iometer logical hard drive test before and you already have iobw.tst files in this section, you'll have to delete them before you do the next test, or you can't change the size of the Iobw.txt file that you tested before. (That means the accuracy of the next test will be affected)

3) Maximum disk size/starting disk Sector/# of Outstanding IO:

General mechanical hard disk sector (sector) size is 512 bytes (sector is the minimum hard disk storage unit). If you use the default 0 o'clock, this means that all the sectors that are filled with this hard drive, if you want to measure a fixed size, can be achieved by setting the maximum disk size. Entering 1000 means that a iobw.tst test file of size 1000x512 bytes will be generated before the test. Starting disk sector refers to the beginning of the first few sectors of the hard disk writing. (to emphasize that, if you set the Maximun disk size number is smaller than the current system buffer, the test results are likely to be buffer performance, the speed will burst high, so appropriate to set a larger point.) ）

# of Outstanding I/O refers to the simultaneous IO request (smart people know that this is the role of NCQ), the default is 1, the general household test will not modify the project, only the manufacturers show scores in advertising will be changed to 32来 run the maximum ioPS users.

4 Write IO Datapattern (newer iometer than the 2006 and 2008 versions of the project)

Repeating Bytes: The generated IO data is duplicated, such as the first IO data is full "A", the second batch is full "B", this data model is used by default in Iometer 2008, and is the most suitable for SandForce control compression algorithm, So SandForce is using Iometer 2008 of the data when advertising.

Peseudo random: The buffer is filled with a random data model, and the new random data is regenerated to the buffer after writing, resulting in a different write IO each time. This data model is used by default from the original version to Iometer 2006, and is very unsuitable for sandforce the main control compression algorithm.

Full random: Generates a set of random 16MB size data models to the buffer. The random data model for this set of buffers is used for each write IO. (although it is random data, but because each random data is the same, so the use of the replica compression technology SandForce controller, the equivalent of each time only need to update the index of random data, or can compress part of the data.) ）

1 Transfer Request size (test requested file size)

You want to test the size of the file block, such as 4KB random to fill in 4KB, the measure 128KB continues to fill 128KB.

2 Percent of Access specification (the current rule accounts for the proportions in this test)

If you need to add more than one rule in the test, this can set the scale, such as 50% 4KB random test, while the other 50% do 128KB continuous test.

3 Percent Read/write Distribution (ratio of read/write to test request)

The read/write ratio of the current test rule, such as 66% read at the same time 34% for the write, for example, our operating system runtime in the reading, the background is also written.

4 Percent random/sequential Distribution (random/persistent ratio of test requests)

Random and persistent percentages, such as measuring 4KB random 80% at the same time 4KB lasts 20%.

5) Align I/O on (align I/o to the specified size)

SSD to partition alignment, here is the truth, generally measured SSD is aligned to 4KB. If you set 4KB, you mean that the data for the test is 4KB aligned each time.

6) Set the global display of the rule

All set up, a name, and then finally look at the next no problem can point to confirm the OK below.

、

1 Run time (test times)

You can set the test time if all 0 represents infinity.

2) Ramp up time

We know that if the disk system uses the array card, there will be a very high cache data, in order to ensure the accuracy of the test data, there can be set a certain time, so that the test results will not calculate this part of the time data, get more realistic test numbers.

1) Update Frequency (Seconds) (Score Update frequency/sec)

The frequency of the score update, you can set the update per second or 60 seconds a update and so on.
Results Since (from which period the result begins)
Start of test-because every second of the test results are different, if you choose this option, then Iometer shows the average from the start of the test to the present.
Last of update-if this is the case, then Iometer will display the average of the latest update to the present. (If you update the frequency to 30 seconds, then the result is the average of 30 seconds)

2 Test Results display area (the picture is the default, you can choose the desired results)
Total I/Os the number of IOPS processed per second under the current rule in Second.
Total MBs/Second current rule, per second transmission rate.
Average I/O Response time (ms) current rule, an average of iops processing times.
Maximum I/O Response time (ms) Current rule, maximum iops processing times.
%CPU Utilization (total) under current rules, CPU occupancy rate
Total error count the test error count in the current test.

In fact, there are a lot of test results can be selected, like in-depth study of friends can continue to explore, I here is simply introduced this powerful Iometer software.