Before you begin
- An existing host pool with session hosts
- The pool's region set (schedule times are interpreted in the pool's Azure region timezone)
- An administrator account (the pool's Settings menu is shown to administrators)
Open the autoscale settings
Open a pool and choose Scaling & Intelligence under Settings in the pool's left-hand menu. The Autoscale Configuration card shows badges for Inherited or Override and ON or OFF, plus Live or Polling for the connection to the autoscale engine. Below the configuration you'll find the Schedules table, and Save Changes at the bottom applies both.
Inherit tenant defaults or override
Each pool either inherits the tenant-wide autoscale defaults or uses its own override. While inheriting, the fields show the tenant defaults and can't be edited; click Override Tenant Settings to set pool-specific values, or Use Tenant Settings to go back. Turning autoscale on for a pool that is still inheriting converts it to an override, so the pool keeps its own copy of the values.
Tenant-wide autoscale defaults are managed in Settings → Configuration Matrix (Scaling) and flow to every pool still set to inherit.
Set capacity targets and load balancing
Capacity targets set the baseline Atlas keeps available regardless of live demand: Session hosts running 24/7, the Default Loadbalancer Type (Depth-First packs users onto the fewest hosts; Breadth-First spreads them evenly), Min. number of available sessions (free session slots kept ready), and Min. number of hosts without sessions. Keep the last one at 0 unless you really want empty hosts running and billing.
Optimize for Power-On-Connect
Optimize for Power-On-Connect applies a preset with no always-on or pre-warmed hosts, so Azure starts a host only when a user connects: hosts running 24/7, minimum available sessions, and hosts without sessions go to 0, concurrent starts go to 1, and the load balancer switches to Depth-First. An AUTO CONFIG ACTIVE badge shows while the preset is on. If any of those fields hold custom values, Atlas asks you to confirm before resetting them, and turning the option off restores your previous values.
Azure's Start VM on Connect must also be turned on for the pool: the Power on Connect switch under Settings → Compute & Deployment → Deployment Defaults. Without it, a pool scaled to zero rejects new connections.
Tune concurrency limits
Concurrency limits cap how many session hosts Atlas starts or stops at the same time, each between 1 and 10. They keep large scale-outs and scale-ins under Azure's rate limits.
Add time-of-day schedules
Schedules handle predictable bursts such as the morning login storm. Click Add Schedule and set whether it's enabled, the minimum hosts, minimum free sessions, minimum idle hosts (leave blank to use the pool default), weekdays, From and To times, and load balancer. A schedule applies only within its time range; outside it, the main configuration applies, and autoscale still adds hosts beyond the schedule's minimum if demand needs them. New schedules default to Monday–Friday, 08:00–17:00, one host, Depth-First. The From and To columns name the pool's region (for example East US time): times follow that region's timezone, not your browser's.
A schedule must apply to at least one weekday, so the last selected day can't be deselected.