Power Management Autoscaling in AVD

What is autoscaling?

By default, Azure Virtual Desktop (AVD) session hosts are statically deployed. While powered on, they incur compute and storage costs. Deallocating VMs saves compute costs, though storage expenses persist. Autoscaling provides demand elasticity by expanding and contracting capacity based on user demand to optimize costs.

A scaling plan can apply to multiple host pools, but each host pool uses only one scaling plan. Plans support multiple schedules for varying requirements like weekdays versus weekends. Autoscale ignores drain mode settings and, in pooled scenarios, ignores load balancing algorithms in favor of those defined in the scaling plan.

Azure’s generally available autoscaling is Power management autoscaling, which adjusts available capacity by powering session hosts on and off based on demand.

Configuring Scaling Plans

Admins select their scaling type after providing metadata like name, resource group, and region. Scheduling options differ between pooled and personal scenarios.

Azure Scaling Plan power management selection

The day conceptually includes four periods:

Ramp-up (start of day)
Peak hours (highest usage)
Ramp-down (users logging off)
Off-peak hours (lowest usage)

Pooled Schedules

Admins specify ramp-up start time, desired load-balancing algorithm, “Minimum percentage of hosts (%)” and “Capacity threshold”:

Start time: When ramp-up logic begins
Minimum percentage of hosts: Percentage of session hosts to start during ramp-up/peak hours (e.g., 10% of 10 hosts = 1 VM)
Capacity threshold: Percentage of capacity usage triggering autoscaling evaluation
Load balancing algorithm: Breadth-first during ramp-up (distributing users to least busy hosts)

Scaling plan add a wizard for ramp up

Peak hours use identical settings except load balancing switches to depth-first for efficient resource utilization.

Scaling plan add a wizard for peak

Ramp-down settings include:

Minimum percentage of active hosts to leave powered on during off-peak
Capacity threshold (higher value allowing host power-down)
Force sign-out options with delay time settings

Scaling plan add a wizard for ramp down

Off-peak hours offer only capacity threshold configuration.

Scaling plan add a wizard for off hours

Personal Schedules

Personal pools allow more aggressive management through ramp-up options:

Start VM on Connect capability for keeping hosts powered off until needed
VMs to start controls which machines power on automatically
Don’t turn VMs on setting with Start VM on Connect maximizes cost savings but may delay logins

Personal pools include unique Disconnect and Sign Out settings:

Disconnected for N minutes: take no action, Hibernate, or Shutdown
Logged off for N minutes: same options available

These settings allow time-dependent optimization strategies, though hibernation and shutdown cause login delays when users reconnect.

What next?

In a follow-up article, we’ll cover Dynamic autoscaling, Azure’s preview capability that goes beyond power management by creating and deleting session hosts.