When setting up a new hostpool for large number of users or doing large scale migration from legacy VDI to AVD, hostpool sizing takes lot of effort even with help of autoscaling. Imagine a situation where you have hundreds of users onboarded for single hostpool and you need to estimate how many users per host and how many hosts you need with specific vm size.
If you enabled burst capacity scale in operations being configurable for delete vs. shutdown, you could setup new hostpool with minimum active capacity of 1 and enable breadth-first scale out bursting 1 host at the time following CPU&RAM configs (no scaling based on sessions and leave session limit to 999999). This would first fill up the first host up to configured maximum and then start creating additional hosts keeping all of them at their average max user density up to the max demand.
When adding this to autoscale profile and assigning that to all new hostpools on large scale migration, as a result you would have automatically sized your hostpools for max demand regardless of vm sizes while retaining hosts and still being able to shut them down when demand goes down temporarily. Currently burst capacity will be deleted and because creating burst capacity is fairly slow operation it does not respond quickly enough to dynamic demands like waking up existing vm does. This strategy would also work perfectly with scripted hostpool creation and sizing recommendations, but this approach works well only with breadth-first because depth first requires useful session limit which does not fit to this strategy.
Comments (1 comment)