M365Con 2027 – free online conference · 11–22 Jan 2027 · 120 sessions m365con.net

Blog · Mirko Peters

Azure Virtual Desktop scaling plans explained: schedules, phases and host pools

Partner · Neutral partner: Nerdio

By ·

An Azure Virtual Desktop scaling plan holds schedules, and each schedule splits the day into ramp-up, peak, ramp-down and off-peak. The plan is assigned to host pools of one type, and autoscale acts on the phase settings.

An Azure Virtual Desktop scaling plan is an Azure resource that contains one or more schedules. Each schedule divides the day into four phases, and autoscale uses the settings of the current phase to decide when to turn session hosts on, consolidate sessions and turn hosts off. This article explains the objects, the four phases, the settings of each phase and the points that decide whether a plan behaves as expected.

Key points

  • A scaling plan is an Azure Resource Manager object. It can be assigned to several host pools, but only to pools of the same type, and the type cannot be changed after creation.
  • Every schedule has four phases. In a pooled schedule each phase has its own start time and load balancing; capacity threshold and minimum percentage of hosts are set in ramp-up and ramp-down.
  • Autoscale turns hosts on when used capacity exceeds the capacity threshold, and it consolidates sessions in ramp-down. Force sign-out applies only in ramp-down.
  • Learn says the Desktop Virtualization Power On Off Contributor role should be assigned with the subscription as scope. A lower scope can stop autoscale from working properly.
  • Hosts that carry the exclusion tag are not started, stopped or put into drain mode by autoscale, but they still count toward the minimum percentage calculation.

The objects: scaling plan, schedule, host pool

A scaling plan is an Azure Resource Manager object that defines the schedules for scaling the session hosts of a host pool. A schedule is a sub-resource of the plan. For a pooled host pool, a schedule specifies the start time, capacity threshold, minimum percentage of hosts, load balancing algorithm and other settings for each phase of the day. For a personal host pool, it specifies the start time and the operation to perform depending on the session state, signed out or disconnected.

One plan can be assigned to several host pools, but each host pool can have only one scaling plan. Learn adds two limits: the host pools must be of the same type as the plan, and the type cannot be changed after creation. A mixed fleet therefore needs one plan per pool type. The plan runs in the time zone you select on the Basics tab of the creation wizard. The scaling plan configuration data must sit in the same Azure region as the host pool configuration. The session host VMs can be in any supported region.

Learn names two scaling methods. Power management autoscaling powers VMs on and off, works with standard session host management and covers pooled and personal host pools. Dynamic autoscaling can also create and delete session hosts, and it applies only to pooled host pools that use session host configuration. Microsoft announced general availability of enhanced host pool management, including dynamic autoscaling, session host configuration and update, and ephemeral OS disks, in July 2026 (Tech Community). Some Learn pages, for example the scenarios and glossary pages and the module version 5.4.2-preview, still carry preview wording; check the current page before relying on a limit. The create-and-assign page lists the Az.DesktopVirtualization PowerShell module version 5.4.2-preview or later for dynamic autoscaling, and the scenarios page says it is not available in Azure Government. Power management autoscaling is generally available. This article concentrates on power management. See the entity pages for power management autoscaling and dynamic autoscaling.

The four phases

Learn defines the phases as follows:

  • Ramp-up: usually the start of the workday. Active sessions increase quickly but have not reached the day's maximum.
  • Peak: the host pool reaches its maximum number of active sessions. The count holds steady and new sessions are still established, usually more slowly than in ramp-up.
  • Ramp-down: usually the end of the workday. Active sessions decrease quickly.
  • Off-peak: the pool usually reaches its minimum number of active sessions. Few users are active, and a small amount of capacity may stay on for people who work after the peak.

In the portal each phase is configured by its start time. Learn adds that on days you do not select for a schedule, the last off-peak settings carry over until the next ramp-up. See also autoscale schedule phases.

Settings per phase for a pooled host pool

In the Azure portal the Schedules tab for a pooled plan asks for these values:

PhaseSettings
Ramp-upStart time, load balancing algorithm, minimum percentage of hosts, capacity threshold
PeakStart time, load balancing algorithm. The capacity threshold carries over from ramp-up
Ramp-downStart time, load balancing algorithm, minimum percentage of hosts, capacity threshold, force logoff of users
Off-peakStart time, load balancing algorithm. The capacity threshold carries over from ramp-down

Capacity threshold. Available capacity is the maximum session limit multiplied by the number of available session hosts. Used capacity is the number of active and disconnected sessions divided by that figure. If used capacity exceeds the threshold, autoscale turns on hosts until usage falls below it. For example, two hosts with a maximum session limit of 20 give a capacity of 40. At a 75 percent threshold, more than 30 sessions trigger a third host.

Minimum percentage of hosts. This is the lowest percentage of all session hosts in the pool that must be turned on in a phase. The result is rounded up to a whole number. In the scenarios page's first example, 30 percent of six hosts is not a whole number, and autoscale rounds up to two hosts. See minimum percentage of hosts.

Load balancing algorithm. For pooled pools autoscale ignores the algorithm configured on the host pool and applies the algorithm from the schedule. The FAQ gives breadth-first for ramp-up and peak and depth-first for ramp-down and off-peak as its example, so that sessions spread out early in the day and consolidate later. The create-and-assign page is narrower: it recommends breadth-first for ramp-up and depth-first for off-peak, lets you pick either for peak, and gives no recommendation for ramp-down.

What autoscale does in each phase

The scenarios page describes the behavior for pooled pools:

  • Ramp-up: autoscale turns on hosts to meet the minimum percentage of hosts and whenever used capacity is above the threshold. It does not turn hosts off in this phase, to avoid a poor user experience.
  • Peak: autoscale turns hosts on and off as needed to keep use stable. It turns a host off when used capacity is below the threshold, turning it off would not exceed the threshold, and the host has no user sessions. At exactly the threshold, autoscale does nothing (FAQ).
  • Ramp-down: autoscale consolidates sessions and deallocates excess hosts, but only hosts that have no sessions, or any host if force sign-out is enabled. With force sign-out, autoscale selects the host with the fewest sessions, puts it into drain mode and notifies users before it ends their sessions. Without it, autoscale waits for users to sign out.
  • Off-peak: autoscale keeps the minimum percentage of hosts running, even if they have no users. The create-and-assign page lists the minimum percentage field only for ramp-up and ramp-down, so check which value your plan carries into off-peak. With 0 percent and no sessions, no host needs to stay on (FAQ).

For consolidation Learn gives this formula: required hosts equal total sessions (active plus disconnected) divided by the product of the maximum session limit and the capacity threshold, rounded up. With eight disconnected sessions, a maximum session limit of 10 and a threshold of 80 percent, the result is 8 / (10 × 0.8) = 1 host.

Force sign-out and wait time

Force sign-out is a setting of the ramp-down phase of a schedule and applies only there. Autoscale drains the host with the fewest sessions, notifies users, waits for the configured time, and then ends the remaining sessions and deallocates the host. Without it, the create-and-assign page tells you to select VMs have no active or disconnected sessions or VMs have no active sessions for ramp-down, and autoscale waits for users to sign out. The FAQ adds that with the "no active sessions" setting autoscale still treats disconnected sessions as active, because users may reconnect immediately. Reaching zero running hosts after hours requires a minimum percentage of 0 percent and no sessions at all; the FAQ suggests the Group Policy setting "Set time limit for disconnected sessions" for that. The FAQ's example: with a 15-minute ramp-down phase and a 20-minute wait, the schedule moves to off-peak first and users are not signed out.

Personal host pools

A personal schedule works differently. The ramp-up phase offers a choice of which VMs to start: assigned VMs only, assigned and unassigned VMs, or no VMs at the start time. Learn highly recommends Start VM on Connect if you do not start hosts in ramp-up. Each phase has a disconnect setting and a sign-out setting. They take a number of minutes from 0 to 360 and an action: deallocate the VM, hibernate the personal desktop, or do nothing. The FAQ adds that autoscale cannot force users to sign out of a personal desktop, and that if hibernate is chosen but not enabled on the VM, autoscale neither hibernates nor deallocates it.

Hibernation needs to be enabled on the session hosts. Learn states that FSLogix and App Attach do not support hibernate, so do not enable it where either is used on personal host pools. See AVD hibernation.

Create and assign a plan

The full portal procedure has its own article, "How to create and assign an AVD scaling plan step by step". In short: set a MaxSessionLimit on every pooled host pool instead of keeping the default, and assign the Desktop Virtualization Power On Off Contributor role to the Azure Virtual Desktop service principal or a managed identity with the subscription as scope. Then, in the Azure portal, search for Azure Virtual Desktop, select Scaling Plans, then Create, and work through the Basics (including the exclusion tag, for example excludeFromScaling), Schedules and Host pool assignments tabs. Changes to an assigned plan apply immediately. To create the plan with PowerShell, the page requires the Az.DesktopVirtualization module version 4.2.0 or later for power management.

Exclusion tags and drain mode

A VM with the exclusion tag is not turned on, turned off or changed in drain mode by autoscale, in any phase. Tagged hosts still count toward the minimum percentage calculation. Learn warns that if every available host is excluded and none can be turned on, new users get a "No resources available" error. The Learn FAQ and the scenarios page both state that for pooled host pools autoscale overrides drain mode on the VMs it scales, whoever set it, so use the tag when you patch or test a host.

Checks after assignment

  • Confirm the role assignment scope. The create-and-assign page says a scope lower than the subscription prevents autoscale from working properly; the troubleshooting page is less strict and lists host pool, resource group or subscription scope, so follow the create-and-assign page.
  • Do not combine autoscale with other scaling scripts. Learn says you cannot use autoscale together with scaling based on Azure Automation and Logic Apps on the same host pool.
  • A host pool needs at least one VM for autoscale to work, and autoscale cannot be set up with the Azure CLI, according to the FAQ. It also cannot ramp down on specific dates such as holidays.
  • Turn on diagnostics: on the scaling plan, open Diagnostic Settings, add a setting, select Autoscale logs, and archive to a storage account in the same region as the plan or stream to an event hub. The records include operationName, resultType and correlationId. For pooled pools Learn recommends the autoscale data in Azure Virtual Desktop Insights for a fuller view.

Related

Sources

Last checked 17 November 2026.

Follow the showApple PodcastsSpotify

FAQ

How many phases does a scaling plan schedule have?

Four: ramp-up, peak, ramp-down and off-peak. Microsoft Learn describes them as the start of the workday, the period with the most active sessions, the end of the workday, and the period with the fewest active sessions.

Can one scaling plan cover pooled and personal host pools?

No. Microsoft Learn says a scaling plan can be assigned only to host pools of the type set in the plan, pooled or personal, and the type cannot be changed after the plan is created.

Does autoscale respect the load balancing algorithm I set on the host pool?

For pooled host pools, Microsoft Learn says autoscale ignores the configured algorithm and applies the algorithm set for the phase in the schedule. The FAQ gives breadth-first for ramp-up and peak and depth-first for ramp-down and off-peak as an example.

Entities in this article

Topics of this article:

AVD scaling plan – Part of Azure Virtual Desktop Azure Virtual Desktop – Cloud platform by Microsoft

Related articles

More articles on AVD scaling plan:

RSS