topic Re: HP P2000 SAS performance in HPE EVA Storage

HP P2000 SAS performance

Ladislav Prokopny — Mon, 30 May 2011 05:39:17 GMT

Hi,

I have 6x 1TB SAS drives in RAID6 and connected to 2 hosts that have HP SC08Ge adapters. What is the typical performance in this scenario ? I'm especially worried about writing, it gives me 80MB/s on writing, but reading is ok. Host is running XenServer5.6sp2 with multipathing enabled.

Is this value ok, or do I have to search for some problem with the hardware/drivers ?

Thank you.

Re: HP P2000 SAS performance

Johan Guldmyr — Thu, 07 Jul 2011 15:12:06 GMT

Hey,

http://h30499.www3.hp.com/t5/Storage-Area-Networks-SAN/HP-StorageWorks-P2000-G3-MSA-Dual-Combo-Controller-FC-iSCSI-LFF/m-p/4779184#M48782

compared to that guy with 8x300GB 15K SAS you are getting good performance.

The 1TB SAS drives are about the same performance wise as SATA drives, as they are 7200 RPM.

HP StorageWorks P2000 1TB 6G SAS 7.2K LFF (3.5-inch) Dual Port MDL Hard Drive

RAID6 is the slowest of the RAIDs when it comes to writing.

Re: HP P2000 SAS performance

Ladislav Prokopny — Mon, 30 May 2011 06:21:09 GMT

Would it be much better when I would change to RAID5 ? What writing performance could I expect then ? (approx.)

Re: HP P2000 SAS performance

Johan Guldmyr — Mon, 30 May 2011 06:26:55 GMT

I don't have an example, maybe somebody else with the P2000 can give us some figures.

As for that, how are you measuring the performance?

Re: HP P2000 SAS performance

Ladislav Prokopny — Mon, 30 May 2011 06:30:07 GMT

reading: 400MB/s hdparm -t /dev/...
writing: 80MB/s dd if=/dev/zero of=/dev/... bs=1M count=1000

and it does not differ when running in the hypervisor or guest.

Re: HP P2000 SAS performance

Johan Guldmyr — Mon, 30 May 2011 06:33:09 GMT

Look into the software iometer to do a little better performance testing.

hdparm is meant for local disks and dd does not reflect real world usage.

Re: HP P2000 SAS performance

Ladislav Prokopny — Mon, 30 May 2011 07:30:58 GMT

Results of 16K, 50% read, 0% random.