Azure Virtual Desktop: Reserved Instances or Scaling
A while ago, I wrote a blog post about saving money by smartly switching the disk types of Azure VMs. But what's about using reserved instances instead of switching VMs on or off (autoscaling)?
I used the same data as in the previews blog post to figure out the best amount of reserved instances for this specific customer environment.
The charts below show the data of a host pool for seven days. Less load on the weekend and more than 400 user sessions on the weekdays:
The maximum number of running hosts was 89 on Thursday.
The VM types are D4 v4, and the current costs in west europe for one hour are:
Pay-as-you-go: € 0.201315 Reserved (1y): € 0.121219 Reserved (3y): € 0.078055
I calculated the costs for this week for different numbers of reserved instances: From 0% to 100% in 5% steps. The results are surprising:
The best value is with 5% reserved instances: € 981.77
The best value is with 25% reserved instances: € 886.42
So, in this case, 25% reserved instances (23 hosts) are an excellent way to go. But it can change from customer to customer even if people work in a shift.