HBA Queue Depth when going to an MSA1000

[Adam Stahl]

Hello all,

I have been trying to tweak my storage a bit and came across many posts that suggest to increase the queue depth from 32 to 64 on Qlogic HBAs. This is straight from VMware support:
"64 on the queue depth is more than what the controller on the MSA 1000 can handle, we will basically be doubling the latency at that point. I provided the option to change it as you requested it. You can feel free to change it if you would like, but it would be against our recommendation and HP's current recommendation."

Is what they say true? I have seen many forum posts that suggest to increase the depth to 64. Does HP not recommend doing that?

[marsh_1]

adam,

i'd agree with the vmware support statement - if you've got issues round throughput on your msa1000, increasing the queue depth will only queue the data up further back up the chain.

fwiw

[Amar_Joshi]

Adam,

QueueDepth setting mostly depends upon the configuration and applications. There is no point agreeing or dis-agreeing with VMWare support, it may be true when you have large configuration on entry-level classified MSA1000 storage. If you have many LUNs presented from MSA to various hosts or various VMWare hosts (not guests) and using OLTP apps, probably you are good with a QD of 32 and further increase of it may give you a problem. And vise-versa for a small setup with less than 15-20 LUNs assigned to hosts without many small IOs may allow you to go for highter QD.

Key point here is that an MSA1000 can handle 41,000 IOPS but depending upon the apps and number of hdds it may be a lot lower. So unless your environment is fully known to someone it's not possible to comment straight on it.

Hope this helps.

Please see the document from Qlogic explains QD and VMWare.

http://www.qlogic.com/SiteCollectionDocuments/Education%20and%20Resource/whitepapers/whitepaper2/FC0054602-00.pdf

[Adam Stahl]

Thank you both for your help and thanks for that link. Good info there.