Explanation
Why the waste happens and who it affects.
Which one is cheaper depends on how many hours each user actually spends connected in a month. Organizations usually pick one mode for everyone at provisioning time and never revisit it, so part-time users, contractors and occasional users sit on AlwaysOn, while full-time users on AutoStop accumulate enough hourly charges to exceed the monthly price.
The mismatch is also created by AutoStop WorkSpaces that do not stop. AutoStop only stops a WorkSpace after it is disconnected, and a client device that is merely locked or sleeping, or a third-party client that the service cannot track, can keep the session connected and billing by the hour. Unlike an inactive WorkSpace, these WorkSpaces are in use and should be kept, only billed differently. AWS offers the Cost Optimizer for Amazon WorkSpaces solution to convert each WorkSpace to the more cost-effective billing option based on its usage.
Billing model
The pricing dimensions that drive this cost.
The running mode determines whether a WorkSpace is billed monthly or hourly.
- AlwaysOn
- A fixed monthly fee for unlimited usage, best for users who use the WorkSpace full time as their primary desktop
- AutoStop hourly usage
- Billed by the hour while the WorkSpace is running, with billing stopping as soon as the AutoStop time after disconnection expires
- AutoStop fixed fee
- A fixed monthly infrastructure charge that continues while the WorkSpace exists, even when it is stopped
- Mode changes
- The running mode can be switched between AlwaysOn and AutoStop at any time
How to detect
5 checks to find it in your estate.
- For each WorkSpace, total the hours a user was connected in recent months using the AWS/WorkSpaces UserConnected metric (per WorkspaceId, filterable by RunningMode) and compare them with the break-even hours for its bundle on the WorkSpaces pricing page
- Flag AlwaysOn WorkSpaces whose monthly connected hours are consistently well below the break-even point, and AutoStop WorkSpaces whose hours are consistently above it
- Find AutoStop WorkSpaces that rarely reach the Stopped state despite short connected time, which indicates clients left locked or sleeping rather than disconnected, or use of third-party clients that cannot signal disconnection
- Review AutoStop Time settings; a long AutoStop time adds billed hours after every disconnection
- Run the Cost Optimizer for Amazon WorkSpaces solution, which analyzes usage data across accounts in AWS Organizations, to produce per-WorkSpace recommendations
How to fix
5 ways to remove the waste.
- Switch low-usage AlwaysOn WorkSpaces to AutoStop and heavy-usage AutoStop WorkSpaces to AlwaysOn with modify-workspace-properties or the console; the running mode can be changed at any time
- Deploy the Cost Optimizer for Amazon WorkSpaces solution to convert billing modes automatically based on each WorkSpace's usage, and review its results before enabling automatic changes in production
- Ask AutoStop users to disconnect or quit the WorkSpaces client at the end of each day, and use the official WorkSpaces clients so disconnection can be detected
- Shorten the AutoStop Time where users accept a quicker stop; the default is 1 hour after disconnection
- Keep AlwaysOn for WorkSpaces where stopping would lose state, such as GPU-enabled or GeneralPurpose.4xlarge and 8xlarge AutoStop WorkSpaces that do not support hibernation, or WorkSpaces with nested virtualization on the Windows versions AWS lists
Documentation
Vendor references for pricing and configuration.