I am looking to find a virtual appliance to test the performance of my Dell EqualLogic PS6000XV SAN.
Ran some tests using Citrix XenServer with their virtual appliance but am now looking for an ESX / VMware equivalent.
Can someone point me in the right direction?
If you just want something simple, make a Windows VM and put IOmeter on it. You can use that for basic benchlining of the system from one VM.
Testing SAN performance is not as straightforward as one would think - but this might be a way you could use to start.
Thanks for the quick replies....
I might go down the IOmeter route, the DVDstore does seem good but a bit long winded to set up....
Do you recommend I put the Windows VM on the local storage and create a volume on the SAN as well and run IOmeter on that?
With Windows iometer can give good info on the storage.
With Linux it depends, and in some cases it is not reliable.
I'm about 75% of the way through building a Windows Server 2008R2 VM now, I'll stick IOMeter on it and let you know the results...
Is there a good general setting to use with IOMeter?
Depends on what you're looking for.
IOmeter will only check against what and how you tell it. Say you make a VM on your local, and give it two hard drives - one with the VM on the local datastore, the other on your PS6K. You can run the check against both 'storage paths' to see the difference in one fell swoop without having to move VMs around. If you put the VM OS disk on the SAN datastore and not local, your IOmeter check will be contending with VM access as well while you measure the SAN, potentially 'flavoring' your results. Of course, the same argument can be posed in the local datastore configuration to a much smaller degree. Either way, the VM storage data should not be of significant impact to the results, but it's a great example of situational awareness in storage environments.
It really depends on what data you're after. I personally use IOmeter to generate a load and resxtop from a vMA to see the HBA/storage characteristics. I like IOPS throughput as the metric I start from. Then I can hand out different load characteristics depending on what I expect the cluster to be doing - 60/40 random read/write, 100 read, 100 write, etc. You know what you're looking for best. Tailor the test to what you expect the load to be and dial it in from there. Remember to do it consistently is the last of my advice.
Good luck trooper.