GRID desktops and memory usage alarms... is this normal behavior?

I have 3 ESXI hosts and 3 Tesla M10 cards. One card in each host. I am using the M10-1B profile. I am using Horizon View 7 to manage the VM’s. I have 30 VM’s spun up and they always giving memory warning alerts the vSphere Client. None of the VM’s are in use (still in testing) so I know they can’t all be using memory. Sometimes the warnings go away and everything is back to normal and sometimes not. Sometimes all the VMs are giving the memeory… sometimes just a few. Odd behavior.

Is this the expected behavior when using a GRID card? Or is something configured incorrectly on my hosts?

This is expected behavior when using ESXi due to the way in which it currently allocates RAM, so don’t worry, it’s fine, although the alerts are annoying.

This is also one of the reasons to avoid ESXi if you’re using other VDI solutions from Horizon, as the licensing works out to be extremely expensive due to 100% of the RAM being allocated to each VM with a GPU, even if it isn’t all being used! … Obviously this won’t affect you though :-)

Thank you for your reply!

Interesting point in regards to using ESXi. Out of curiosity what would your recommend solution be?

Lots of variables to consider…

Everything below is all just my opinion. Other peoples opinions will obviously vary depending on their experiences.

  • Are you using Horizon, XenDesktop or something else
  • Are you using GPUs
  • What’s the workload
  • Partner status
  • Existing infrastructure
  • Local skills / product familiarity (Design, Implementation, Support) etc etc.
  • Product integration and compatibility with existing technologies (For example, does your security, monitoring or management solution work with X, Y or Z)
  • Does it do what you need
  • Future technology road map and eco-system
  • Functionality
  • Usability
  • Cost
  • etc etc etc

There’s lots to consider.

I’d like to take the best bits out of ESXi and XenServer and combine them, as they both have advantages over each other.

  • XenServer licensing is far far better than ESXi.
  • XenServer GPU management is excellent and ESXi basically doesn’t have any.
  • XenServer is a lot cheaper overall due to the licensing.
  • ESXi has a much better, more modern user interface.
  • ESXi has a lot more options in terms of security and overall management.
  • ESXi has better PCIe management
  • etc etc etc

They both perform very well from my experiences and use cases, although I dare say that in some scenarios one will be a better choice than the other.

They both have positives and negatives and ultimately it depends on a variety of considerations.

Hyper-V is getting there, but it’s not something I see much of so can’t really comment on it, however I’m hearing good things about the most recent version.

I am also having this same problem, just on a much larger scale. This so far appears to be a cosmetic issue.

BJones:
"This is also one of the reasons to avoid ESXi if you’re using other VDI solutions from Horizon, as the licensing works out to be extremely expensive due to 100% of the RAM being allocated to each VM with a GPU, even if it isn’t all being used! … Obviously this won’t affect you though :-)"

What do you mean by this exactly? How does licensing affect RAM? Are you speaking of vRAM?