How to run the IOZONE benchmarks

[Dixit]

So Ive read the "How We Test Networked Storage Devices - Revision 3" which is linked here: http://www.smallnetbuilder.com/nas/nas-howto/30682-how-we-test-networked-storage-devices-revision-3

Now the command they use is this:
iozone -Rab [results file name] -i 0 -i 1 -+u -f [path to directory on NAS device under test] -y 64k -q 64k -n 64k -g 4G -z

Now I basically run the exact same command but with the path to my HP Mediasmart and changed it to 1G size instead of 4G.

Now I cant get anything near the results they they show in their charts and what not. Here is an example of my output, the whole WRITE report makes absolutely no sense to me.

====================================
iozone -Rab 1gigtest.xls -i 0 -i 1 -+u -f //hpwhs/Public/test -y 64k -q 64k -n 64k -g 1G -z

The top row is records sizes, the left column is file sizes
Writer Report
64
64 122797
128 123858
256 165706
512 251066
1024 346702
2048 584585
4096 697248
8192 817579
16384 609229
32768 1005866
65536 1034785
131072 1123276
262144 1098445
524288 1147048
1048576 288798
Re-writer Report
64
64 1704571
128 2226061
256 2613980
512 3000045
1024 3078184
2048 3142674
4096 3229863
8192 3214224
16384 2887863
32768 2964063
65536 2977322
131072 3172291
262144 3923570
524288 4573309
1048576 389992
Reader Report
64
64 2366541
128 4642928
256 159368
512 119567
1024 92398
2048 92358
4096 93626
8192 91788
16384 94379
32768 93765
65536 79277
131072 87362
262144 86502
524288 65724
1048576 79948
Re-reader Report
64
64 3207285
128 5603533
256 142943
512 119604
1024 89289
2048 96260
4096 88340
8192 94524
16384 94304
32768 86157
65536 67514
131072 94436
262144 84431
524288 75983
1048576 88807

Writer CPU utilization report (Zero values should be ignored)
64
64 0
128 50.54659271
256 0
512 3.47009E-15
1024 0
2048 0
4096 17.3778801
8192 8.83159256
16384 3.121222258
32768 4.652143002
65536 2.889974356
131072 3.019198656
262144 2.593460798
524288 2.582196474
1048576 2.604515791
Re-writer CPU utilization report (Zero values should be ignored)
64
64 0
128 2.03886E-14
256 0
512 45.13013077
1024 0
2048 0
4096 0
8192 0
16384 1.05102E-14
32768 1.660509706
65536 1.779135466
131072 1.863033772
262144 1.398760438
524288 1.178601384
1048576 1.778490663
Reader CPU utilization report (Zero values should be ignored)
64
64 0
128 4.23417E-14
256 0
512 0
1024 0
2048 0
4096 0
8192 0
16384 2.99059E-14
32768 0
65536 1.732319355
131072 4.112999916
262144 3.03336215
524288 3.275306225
1048576 3.310921192
Re-reader CPU utilization report (Zero values should be ignored)
64
64 0
128 5.12606E-14
256 0
512 0
1024 0
2048 0
4096 30.95363998
8192 0
16384 3.02877E-14
32768 3.86485219
65536 4.784247398
131072 3.17681E-15
262144 4.014090538
524288 3.151641846
1048576 2.885022163
====================================

If you look at my attachment and look at the WRITE report, it makes no sense. This is a gigabit connection between the test machine my HP Mediasmart 485 also gigabit. I cant get anything near the same chart or info shown in the how to test link.

Any help would be appreciated.

Dixit

 

[wolverex]

The results are in kB/sec which means some of your results even exceed 1 GB/sec which is surely not possible with the Gigabit switch you have, I think this is what you mean by not being able to get the correct results?

Are you sure you have pointed to the right path? Yours used //computer/share but i think it should be \\computer\share ? (i'm not sure about the syntax IOzone uses)

When running check your storage device to see that it is in fact in activity and with hard disk activity.

I ran mine just a short while ago with no problem. Maybe it does not accept the \\computer\share\ input? What i did was map the network share to a drive on my com and point -f to Z:\test
The path is not exactly a path but a filename that will be used temporarily for the scan. You should do read and write together as the write process creates the file for the read process to work.

 

[00Roush]

This might give you a little more info about the higher write results you are seeing... http://www.smallnetbuilder.com/content/view/30769/89/ If you want to know more I can explain in depth why this happens for writes.

As for the path... it depends on if he is running iozone from windows or linux/unix.

Basically to get more accurate write numbers you need to test with file sizes that are larger than the amount of memory you have in the client machine.

Based on you results though it looks like reads average around 80-90 MB/sec which is pretty good.

00Roush

 

[Dixit]

The results are in kB/sec which means some of your results even exceed 1 GB/sec which is surely not possible with the Gigabit switch you have, I think this is what you mean by not being able to get the correct results?

Are you sure you have pointed to the right path? Yours used //computer/share but i think it should be \\computer\share ? (i'm not sure about the syntax IOzone uses)

When running check your storage device to see that it is in fact in activity and with hard disk activity.

I ran mine just a short while ago with no problem. Maybe it does not accept the \\computer\share\ input? What i did was map the network share to a drive on my com and point -f to Z:\test
The path is not exactly a path but a filename that will be used temporarily for the scan. You should do read and write together as the write process creates the file for the read process to work.

//computer/share is the proper syntax when using the command on windows, if you use the \\computer\share it will error and tell you must use the other method. It is definetely hitting that share as I can see the NAS jamming away.

This might give you a little more info about the higher write results you are seeing... http://www.smallnetbuilder.com/content/view/30769/89/ If you want to know more I can explain in depth why this happens for writes.

As for the path... it depends on if he is running iozone from windows or linux/unix.

Basically to get more accurate write numbers you need to test with file sizes that are larger than the amount of memory you have in the client machine.

Based on you results though it looks like reads average around 80-90 MB/sec which is pretty good.

00Roush

00Roush hit it on the head, I forgot to update this thread but the editor told me about that article you mention, and using the "-c" is what fixed it, its a new algorithm to combat caching and incorrect values.

Dixit

 

[thiggins]

00Roush hit it on the head, I forgot to update this thread but the editor told me about that article you mention, and using the "-c" is what fixed it, its a new algorithm to combat caching and incorrect values.

Cached performance is important to understand if you are interested in understanding write performance at smaller file sizes. Why Cache Matters in NAS Performance
I always just map the share on the NAS that I'm going to test. Much easier than futzing with paths.

 

[Dixit]

Cached performance is important to understand if you are interested in understanding write performance at smaller file sizes. Why Cache Matters in NAS Performance
I always just map the share on the NAS that I'm going to test. Much easier than futzing with paths.

No doubt there. I was initially just thrown off on how the values from IOZONE came in at way above the gigabit threshold.

Dixit