Within the FinOps Framework, rate optimization is one of the capabilities developed to optimize the use and cost of cloud environments. It is the ability to reduce the cost/hour of a resource by negotiating lower rates. While usage optimization focuses on the strategy of using fewer resources, cost optimization focuses on the strategy of paying less for the resources used.
Azure Reservations are commitment-based discounts, where Microsoft reduces the cost of a family of machines by up to 72% for users who commit to using the contracted instances for a period of 1 or 3 years. It is an excellent cost-reduction strategy for resources that are used consistently and continuously.
Machine Families – Virtual machine size flexibility
In general, Azure reservations have flexibility, so VMs in the same flexibility group can benefit from the coverage.
If I purchase an Instance Reservation for SKU Standard_DS3_v2 (Ratio 8), and I happen not to have a machine of that size, the following machines, if they exist in the environment, can benefit:
- Standard_DS1_v2 (Ratio 1): the discount would cover up to 8 machines.
- Standard_DS2_v2 (Ratio 2): the discount would cover up to 4 machines.
- Standard_DS4_v2 (Ratio 4): the discount would cover up to 2 machines.
- Standard_DS5_v2 (Ratio 16): the discount would cover 50% of the costs, and the other part would be paid in the Pay-as-you-go model.
A good practice when performing Rightsizing of machines is to try to keep them in the same coverage family to avoid losing the benefit. If a machine is switched to a different family, the user will have to bear the costs of the machine in the Pay-as-you-go model and the costs of the reservation, which is charged monthly.
Microsoft itself provides the list of families and coverages at the following link.
Reservation Management
- Billing Account Administrator
- Enterprise Agreement Billing Administrator
- Microsoft Customer Agreement Billing Administrator
Is it possible to return Reservations?
Microsoft allows the return of up to USD 50,000.00 within a 12-month period. This amount applies to the total sum of all returns and considers the total value of the 12 or 36-month contract.
Example:
I have an instance reservation for 5 machines, lasting 3 years, valued at USD 5,000.00. After 18 months of the contract, I decide to delete the resources that would have coverage. If I wish to return the reservation, the limit is calculated based on the remaining time of the contract, which is 18 months.
USD 5,000.00 * 18 = USD 90,000.00
In the above scenario, the reservation could not be returned. However, if I wish to return only a part of the instance reservation, equivalent to 2 machines, the scenario would be as follows:
USD 2,000.00 * 18 = USD 36,000.00
Yes, it is possible to make partial returns of instance reservations when the contract corresponds to more than one machine. In the above scenario, I could not only return part of the reservation but also have a remaining balance of USD 14,000.00 for future returns.
In practice, my suggestion is to return reservations only as a last resort. The ideal is to exchange for other machine families to avoid losing the return benefit in more critical scenarios, such as migration to another cloud.
What are the rules for exchanging Reservations?
- Reservations can only be exchanged for others of the same type;
- Reservations can be exchanged for others in a different region;
It is important to consider that exchanging the reservation restarts the 12 or 36-month period, regardless of how much time was left on the previous reservation.
Recently, Microsoft changed some rules regarding the returns of Instance Reservations, which I addressed more specifically in this article.
Which Reservations cannot be returned?
- Azure Databricks
- Synapse Analytics Pre-purchase plan
- Azure VMware solution by CloudSimple
- Azure Red Hat Open Shift
- Red Hat Plans
- SUSE Linux Plans
In conclusion…
Based on the information above, you are fully equipped to purchase Azure Instance Reservations. Use the benefit wisely, and always consider the real needs of your cloud environment. Remember that acquiring reservations is the last step in optimizing computing costs, which, unless otherwise advised, should follow this order:
- Start/Stop policies for non-essential resources
- Rightsizing of machines
- Instance Reservations
Check the original article in Portuguese here
