HP P4000 high queue depth

[Martin Hanspeter]

Hello,

i`ve a 2 Node Cluster running on SAN/IQ 10 with 3 ESXi 5 Hosts. There are running ca. 20 VMs.

Once or twice per day it occurs that the queue dept in the Performance Monitor goes over 500 and all VMs remains blocked for ca. 30 seconds. After this 30 seconds the system will run normaly and the queue depht shows ca. 10-15.

There is no backup or some other high I/O operation running during this time.

Thank you

[danletkeman]

Same sort of problem here.  Very poor performance all day long, medium to high queue depth, I have migrated all but three VM's off of it, and still performance is horrible.  I have a VSA san to compare it to, which is running on old ML350 G5 750GB sata drives, and it performs at least 5 times better, running on the same infastructure and using the same esx hosts.

I currently have a support request in, we have gone through everything, but no answer yet.  Its in HP's hands right now.

At this point our san is useless to us.

Let me know if you figure anything out.

Dan.

[Jitun]

For a 2 Node cluster the Qdepth is quite high.
Q depth is determined at the host and is not anything the SAN can change.

Optimal Q Depth depends on the type and number of Drives you have.
For SAS Drive optimal Qdepth is 1.5 to 2
For SATA Drive optimal Qdepth is 1

Depending on whether you have 8 drives or 12 drives per node, one can calculate the Optimum Q depth

So for a P4500 with 12 drives
with SAS drives optimum Q depth is between 18-24 per Node
with SATA drives optimum Q depth is between 12 per Node

So for a P4300 with 8 drives
with SAS drives optimum Q depth is between 12-16 per Node
with SATA drives optimum Q depth is between 8 per Node