Lavinia Spatariu | March 29th, 2022
On Sep 8th 2021, Microsoft announced the public preview of On-demand Capacity Reservations for Azure Virtual Machines. What this means is that organizations now have the option to reserve Compute capacity with a guaranteed SLA without having to make a monetary commitment. Once enabled, the capacity becomes available instantly and it remains dedicated only to your resources, for as long as you need it. On March 28th 2022, On-Demand Capacity Reservations for VMs became generally available. Azure Site Recovery also integrates with capacity reservations (GA on March 28 also).
How is this different from Azure Reservations?
Although both options are available for Azure VMs, there are some key differences between on-demand capacity reservations and Reserved VM Instances (RIs). The two can be combined however for significant cost savings.
From the Announcement page
When should On-Demand Capacity Reservations be used?
Capacity Reservations should be considered every time an organization has compelling capacity needs and workload prioritization is of essence. Some common use cases:
- Anticipated peak capacity events such as a new product launch;
- Mission-critical applications which rely on underlying compute resources;
- Disaster Recovery (and outside DR times, the reserved capacity can be transferred to other resources).
Which VM types does this apply to?
B-series, General-Purpose (Av2, Dv2 and newer generations), memory-optimized (Ev3 and newer) and Compute-optimized (Fv1 and newer). Microsoft is planning to expand the eligible VM range in the future and will be making the on-demand capacity reservation option available to Azure VMs for the US Government as well.
How does it all work?
- Capacity reservations can be started at any time and can also be deleted at any time.
- There is no commitment which means the price of the running VM will still depend on its configuration (size, region etc). You need to select either an Azure region or an Availability zone and a quantity. Once these parameters are defined for the required reservation, you are effectively sending a capacity request which will be accepted or denied (if Azure does not have the specified capacity available). When the reservation is accepted, you can start assigning VMs to it.
- The quantity can be changed later on but the region and the VM size cannot. If you need to change these two properties, Microsoft recommends to delete the existing reservation and create a new one based on the new size and location needs.
How does Billing work and what if I already have Reserved Instances?
- If you currently have 1-year or 3-year RIs in place, the discounts will be automatically applied to the eligible VMs and so reserved capacity and Reserved Instances can be used together. If you reserve 10 VMs and 5 of them are covered by an existing RI, then you only pay for the remaining 5 VMs (until the end of the RI term or until the committed amount is consumed, whichever comes first). More detailed info on billing can be found here.
- Very important: make sure the reserved capacity is actually used once activated because you will start getting billed for the compute resources you reserved straight away, whether you use them or not. If for example you reserve 10 VMs but only deploy 5 VMs with the reserved capacity as a property, you will still incur charges for all 10 VMs.
Are there any limitations?
Capacity reservation cannot be transferred to other Azure subscriptions during Preview and is not available for:
- Free Azure Trials or Azure for Students;
- Availability Sets, Dedicated Azure Hosts, Spot VMs;
- Proximity Placement Group, UltraSSD Storage, Update Domains;
- Deployments with more than 3 Fault Domains (FDs);
How to use on-demand Capacity Reservations for Azure Virtual Machines