testing rw speed for HPUX HD

[haux]

Is there any utility for HPUX 11v2, than can give me the Read/Write speed.
Like the command "hdparm" on Linux.

http://www.justlinux.com/nhf/Hardware/Hard_Drive_Speed_Tweak_for_Linux.html

Thx

[Peter Nikitka]

Hi,

I did not see a port of hdparm ro HP-UX yet, but you can give 'dd' a try.
Read speed can be determined always:

To read 100mb of data, use
time dd if=/dev/dsk/xxxx of=/dev/null bs=1024 count=100

If there is space on a disk not already used for logical volumes, create a new one using (part) this data (e.g. 100MB), do NOT create a filesystem and then try
time dd if=/dev/zero of=/dev/vgxx/rlvolyy bs=1024k count=100

mfG Peter

[The Old Man]

Hi,

A free, comprehensive tool to measure IO at:

www.iozone.org

There's an HP-UX tarball for it....

There is also HP DiskBench (officially unsupported, but a good utility) on hps DSPP website.

Tom

[haux]

Hi Peter,
I made two tests :
1) I created a new lvm on EVA and made a write test:

My_Server1:/>time dd if=/dev/zero of=/dev/vg_test/lvol6 bs=1024k count=100
100+0 records in
100+0 records out

real 1:07.7
user 0.0
sys 1.5

So it writes 100Mb at a speed of (1024*100/67)=1528.358 Kbites/s


2)On another server connected to a disk array with one FiberChannel i get these results :

My_Server2:/ >#time dd if=/dev/zero of=/dev/vg_test/lvol24 bs=1024k count=10
10+0 records in
10+0 records out

real 11.2
user 0.0
sys 0.3

So it writes 100Mb at a speed of (1024*100/11)=9309.09 Kbites/s


So the EVA 8 times less quick than the simple disque array, do think those results are logiques ?

Best regards.

[Alzhy]

Messr Haux,

Is server1 the same as server2 - CPU, HBA, memory, OS, etc?

Also, try doing your dd tests using RAW device: of=/dev/vg_test/rlvol1

And one more thing, make your test size (and test lvol larger --- say larger than the memory of your system)


I would think there should be no difference or only a minute one between the 2 arrays.

The reasons of such a wide disparity that I can think of are:

1.) EVA on server1 is loaded/taxed
2.) Your FC config (HBA, paths, etc.) on server1 is faulty...

[Peter Nikitka]

Hi haux,

your 2nd try used count=10, not count=100 -
Thats a factor of 10.

mfG Peter

[Alzhy]

Ahhh,

One thing for certain -- the EMC on Server2 will simply outgone the EVA for such a measly 100MB dd test. Why? I am dead sure your EMC has Gigabytes of Cache on it whilst the EVA 5K only has about 1GB per controller...

Try a bigger LVOL.. say 10GB... and you;ll be surprised, the gap will narrow...

Again use RAW lvols and also try varying your blocksize...