Recently Microsoft released a new offering called Azure Savings Plan (officially called Azure Savings Plan for Compute) which provides additional discount on compute resources based on a committed spend and term.
If we think about discounts within Azure, there is currently the Reserved Instance offering which provides additional discount based on committing to a term of using a resource, the longer the term the higher the discount typically. Reserved Instances aren’t going away with the introduction of Azure Savings Plan, instead this new offering compliments and works in conjunction with Reserved Instances.
If Reserved Instances aren’t going away, then what is the purpose of the new Azure Savings Plan? If we take a step back and look at Reserved Instances and how they work, we can then build on this to explain the differences as compared with Azure Savings Plan.
Reserved Instances allow you to commit to a 1- or 3-year term (and some offerings for 5-year terms) and either pay up front or monthly for that term, the reserved instance itself once purchased is then locked against a specific resource type, and location. If you decide to move the resource to a different location or change the resource type (one VM type to another for example, like a D4s to a E4s), then the reserved instance you purchased will no longer apply and your resource will default to the pay as you go consumption billing rate until you update the reserved instance to match. Changes to reserved instances is supported to a degree with some limitations around what you can change to and how often, but for most partners the limitations apply less than you might think. You also have the option of cancelling the reserved instance completely if you no longer require it with a potential 12% cancellation fee (of the remainder value) that could apply (as of October 2022 it is not being charged for by Microsoft).
Azure Savings Plan is provided in either a 1- or 3-year commitment term which can be paid monthly or upfront. The commitment term is based on hourly spend and once committed to, it cannot be cancelled (as of October 2022 there are no exceptions). However, the Savings Plan isn’t fixed to specific virtual machines or regions, instead it applies to any virtual machine in any region. Say for example you have a hypothetical bunch of virtual machines which come to $10 an hour in consumption normally, but with a Savings Plan you may receive 50% discount, this would mean you could look to commit to a $5 an hour Savings Plan and it will reduce the virtual machine pricing to that value.
There is some logic behind the scenes as to how Reserved Instances, Savings Plan’s, and standard pricing apply. Effectively Reserved Instances take precedence over a Savings Plan as they will in general have a higher discount applied to them. As Savings Plans can apply to almost any compute resource, the system works out which resource it should apply to based on what will provide the greatest savings and then applies the Savings Plan accordingly, every hour this process is re-run to make sure the best Savings Plan is applied against the appropriate resources until the Savings Plan hourly commit is exhausted at which point the normal pay as you go rate then applies and standard pricing is in place.
When I think about Reserved Instances and the new Savings Plan, I think MSP’s may stick with Reserved Instances for the most part as most of the workloads are static in nature and assigning a Reserved Instance will provide the greatest discount and has an option of cancelling if required. Savings Plan I think will be used more for partners who have customers that have more dynamic workloads where they are turning virtual machines on and off all the time and moving them around regions as Savings Plan provides the greatest flexibility in that regard but at the cost of a commitment term which cannot be cancelled.
If you are after more details on Azure Savings Plan, please check out the below documentation from Microsoft which provides more in-depth information on the subject.
