Performance Tests for VDI

  • 19 March 2014
  • 3 replies
  • 5192 views

Badge +6
I would be interested in seeing some performance tests and results from VMware Planner, Login VSI, etc. What are you doing to load test your VDI environments?

I usually use LoginVSI. Just curious what you use.

3 replies

Userlevel 2
Badge +13
I'm just writing up a powerCLI script to spin up desktop pools and do things like enable/disable shadow clones and grab metrics through the REST API. It's only going to measure provisioning though, not workloads.

I got held up because it seems that even though VAAI is now a supported feature in View 5.3, there is no parameter to enable it when creating a pool. Wierd thing is that everything spins up so fast it looks like it is enabled, even though it says it's disabled if you look at the pool in the View Administrator. I opened a support ticket up with VMware, who said it's not supported yet. I'd appreciate anyone else calling up and saying where's my VAAI support in PowerCLI, maybe it will make it to View 5.4 then.

We've been having issues with refreshing thrashing and the disks getting kicked offline and nodes being thrown out of the metadata ring that is supposed to be fixed in 3.5.3, so I'm really not looking at the script to measure max performance, but can I have 1000 users log off at the same time and not get any nodes thrown out of the metadata ring.
Badge +6
Login VSI here too usually with the medium workload. Not a bad way of at least getting a middle-ground result on the usual stuff end users do.

I tend to add some chaos into the mix by randomly booting/rebooting 20% or so of the estate while the VSI tests are running on the other 80% just to try and replicate the worst scenarios you might see during a day. Generally if you can get away with a boot-storm you can handle anything :D

Maybe mixing the scripting idea above with the regular tests is the way forward?

The vast majority of performance data I've seen on joint Citrix-VendorXYZ testing uses Login VSI too so I think it's a good starting place anyway.


David
Userlevel 2
Badge +12
I always use LoginVSI with the new benchmarking mode in 4.0 which locks all of the variables (based upon medium workload) for apples to apples comparisons.

In case you haven't seen them all of our RAs have LoginVSI test data:
  • http://go.nutanix.com/TechGuide-Nutanix-CitrixXenDesktoponHyper-VReferenceArchitecture_Asset.html
  • http://go.nutanix.com/TechGuideNutanixXenDesktopandvSphereonNutanixReferenceArchitecture_LP.html
  • http://go.nutanix.com/TechGuideNutanixHorizononNutanixReferenceArchitecture_LP.html
I would also note that LoginVSI is good at simulating workload, however I wouldn't use it for sizing as a PoC is always best 🙂

Reply