intelligent CPU & RAM usage autoscale triggers per AVD session host for scaling logic auto-scaling

Thank you for your feedback. Nerdio releases a new version every month with many new features. Please watch the release notes page, https://nmehelp.getnerdio.com/hc/en-us/articles/19837802929677-Release-Notes , every month to see what is coming in the next release. More intelligent capabilities are slated to come out in future releases.

Greg Roberson, does this means that this feature is already on Nerdio's roadmap and that it will be included in one of the next releases?

hi Joeri kumbruck - thanks for this suggestion. We have had similar feedback in the past, and we do have this request captured, however it doesnt currently have a planned inclusion date. We'd also be keen to hear from others if this feature would be valuable! 

Hi Joeri kumbruck thanks for submitting this feature request.
Toby Skerritt this feature would be of extreme value.
We've had situations where the average resource utilization across all session hosts was acceptable, however multiple session hosts were actually flat-lining.
Implementing a load evaluator at a session host level to assist with assessing logon capacity and scale-out needs would significantly help in these cases.
Otherwise we need to over-allocate resources, introducing ineficiencies and additional costs.

I miss the feature too. But as I understand it, this is something that Microsoft should build into the product. After all, sessions are no longer permitted based on the load. This function is based on the VDA agent. The most Nerdio could do here is check whether a load is present and block VMs accordingly (drain mode). Unfortunately, this would always be delayed. The only thing we can already do today is to configure the scaling logic based on CPU and RAM.

Toby Skerritt, do you have any ideas on implementing this without an agent? Every time I've thought about it, I've concluded that it probably can't be done without direct system access.

Stefan Beckmann, yes an agent would be the best approach here. We can detect CPU & RAM load (and we surface this information per host), so it is achievable, but there is a small delay and an API consumption impact here.

Toby Skerritt & Stefan Beckmann, maybe Nerdio should hire an ex-Citrite who is familiar with Citrix's Load evaluator mechanism and do the same trick here ;-) 

Toby Skerritt, yes, I know this possibility. But it's delayed, and in my understanding, it does not work for a load evaluator.

Joeri kumbruck I think Microsoft should hire a few employees from Citrix. As I have already written, this is a functionality of the VDA and, in my opinion, should be in the AVD Agent. But let's see...

Toby Skerritt - I think this feature could be of use, if it can be expanded to include running scripted actions, similar to the auto-heal actions.

Quite old topic but it is still valid. It is quite often then on average host pool from autoscale point of view is fine, but single session host can struggle with load much over the limits. If number of logged users is still lower than on other session host - new session still can end up on this particular session host.

To solve this issue, autoscale calculation metrics should be done also per individual session host, if cpu/ram usage is above the limits, session host should be placed in drain mode - this will stop new session to log in to this session host. 

Host in drain mode should not be included in average cpu/ram consumption. This last step is to avoid unnecessary autoscaling events if other session host consumption is quite fine, and more session hosts actually are not needed to be activated.

any news on this? BTW, you can ask MSFT to have this builtin in the AVD product, but Citrix didn't do this either in the past (Windows Terminal Server) and they built their own (patented?) load evaluator mechanism. What's stopping Nerdio to do the same ;-)? (except for the technical challenge)