Greetings,
I'm looking for some answers and am hoping to find them here...
I have a ~100 persistent Citrix VDI build on 3 ESX 5.5 hosts. The VDIs are Win7 w/ 2CPU, 4GB RAM & up to 128MB vRAM.
There is a single Raid5 datastore (EMC VNX5100) that their vdisks reside on and sometimes performance takes a hit when there's a disk IO storm.
At the recommendation of a consultant, a SSD datastore on the VNX was provisioned and the majority of VDI VMs had their Swapfile location assigned there.
It's been a couple of months since and we've recently experienced another VDI disk IO storm. During the storm I monitored the IOPS (under the performance tab) of the datastores and the one the VDIs had their vdisks on saw significant read IOPS and coinciding increases in latency, while the SSD datastore that the vswp files are on saw near-zero activity.
Under normal operating conditions the SSD datastore also showed near-zero activity.
Is this a case of the vswap files not actually being accessed at all? Or does the access of them not show up in the performance monitoring of vCenter?
If it is the case of the former, then is it because the VDIs haven't required more than their allocation of RAM?
Thanks in advance for any answers and/or clarification!
Cheers,
Andrew
