As with many solutions in the cloud, performance is important. However, performance needs to be balanced with cost. As previously covered, the cost for users in AVD is via licensing, many of which organisations will already be using today. An additional cost comes from the AVD service and infrastructure running in Azure, which includes Sessions Hosts, Storage, Logging, Backups and Networking.

Licencing costs are easier to predict than the infrastructure, but with correct planning and an understanding of the available features you can optimise for costs. Some key areas to think about when balancing cost with performance include:  

• Multi-Session: With AVD you have the ability to spread multiple users on Windows 10/11 hosts using multi-session capabilities. Be sure this meets the requirements of the users and apps. In many cases, multi-session is possible and will save costs, not just on the number of session hosts but also by reducing management overhead.
  
  
• Scaling plan: The Azure Virtual Machine session hosts will contribute to at least 70% of your overall Azure infrastructure costs. Therefore, controlling the number of session hosts will greatly optimise your spend on AVD. Scaling Plans give you the ability to scale session hosts either up or down based on the time of the day, days of the week or session limits. Configuring this correctly will ensure you are providing just the exact amount of compute required and thus optimising costs.
  
  
• Auto-shutdown: For simpler AVD deployments with more predictable working hours, shutting down the Virtual Machines will help with costs. A business requiring a small number of session hosts with a usage time of 9-5 Mon to Friday will benefit from simply switching the hosts off during out of hours. More complex environments would likely need Scaling Plans.
  
  
• Start VM on connect: One handy feature to optimise the user experience alongside the reduction of costs is ‘Start VM on Connect’. This allows users to turn on their session hosts only when they need them. If a user does want to work over a weekend or in the evening, this feature will allow you to schedule shutdown but still provide a method of accessing the desktop. Obviously, an increased connection time is needed as it waits to start the VM but the users are made aware of this when connecting.
  
  
• Right-Sizing: Session hosts sizing and the compute resources (vCPU/Memory) you assign to each user largely depend on workload requirements and specific use cases. However, as with all VMs in Azure, ensure you size them correctly and try to avoid over-provisioning compute resources.
  
  This becomes especially important when delivering personal desktops to your users instead of pooled as the number of session hosts you require to manage and therefore pay for increases significantly. Before deploying, estimate resource requirements based on current usage. Once you have provided a host, use logging and monitoring to understand real-world compute requirements and re-size accordingly.

• Reserved Instances / Savings plans: As with any VM in Azure, Sessions Hosts gain huge cost savings via Reserved Instances or Saving Plans. Reserved Instances (RI) allow you to save by committing to spend over a 1 or 3-year period.
  
  Saving Plans allow the reduction in cost by committing to spending a fixed hourly rate over a 1 or 3-year period. Using the above plans can save up to a whopping 72%. As always with these offers, it doesn’t always stack up on top of other cost-saving optimisations, especially reserved instances. Ensure the plans make sense, they are largely aimed at resources that run 24 hours per day and with no plans to be decommissioned in the next few years. For example, if a VM is assigned an RI you won’t benefit from shutting it down as you’ve already committed to the spend on it running 24/7. RI’s benefit business whereby session hosts are always going to run.