On 9/4/2013 13:44, mcclnx mcc wrote:

for sequence I/O performance test we can use "hdparm -t /dev/xxx"

bur for random I/O performance test which Linux command I can use?

I use Bonnie++ for that:

http://www.coker.com.au/bonnie++/

Actually, I've been using the experimental pre-2.x version:

http://www.coker.com.au/bonnie++/experimental/

It's a far better test for sequential I/O than hdparm -Tt, too.

** our environment does NOT allow install third party software..

Not even if it's:

a) only for a test b) never installed, and c) removed immediately after the test completes?

Bonnie++ is self-contained, easy to build, and will run from the build directory without having to be "make install"ed. You can just remove the build directory when the test is done. There will be no traces that the program was ever present. We won't tell anyone you violated policy. :)

Am 04.09.2013 21:44, schrieb mcclnx mcc:

we just purchase new I/O card and like to check I/O performance.

for sequence I/O performance test we can use "hdparm -t /dev/xxx"

I think hdparm is really sub-optimal. "hdparm -t" will not give you any reliable value. "dd" instead lets you control parameters like size of data transmitted, and has control feature through oflag and iflag, and lets you test read and write spead independently.

bur for random I/O performance test which Linux command I can use?

fio http://freecode.com/projects/fio

Available through EPEL.

** our environment does NOT allow install third party software..

Don't know whether EPEL is already forbidden land for you.

Thanks

Alexander