Ensure failover capacity at optimal cost with Azure Site Recovery

Business Continuity and Disaster Recovery (BCDR) is becoming increasingly crucial as the data estates of customers are growing. Azure Site Recovery (ASR) is the Virtual Machine (VM) Disaster Recovery (DR) service provided by Azure and is a key component of several customers' BCDR strategies. ASR enables the replication of workloads from a primary location to a secondary location. In the event of an outage at the primary site, workloads can be accessed from the secondary location by failing over. Once the primary location is operational again, it is possible to fail back to it. In this article, we address a common question: How can you improve the chances of capacity availability for completing failovers at optimal costs? 

When planning the Disaster Recovery strategy, it’s essential to understand that the cloud ecosystem operates on a shared responsibility model between customers and cloud providers.  If compute capacity is unavailable in the DR region due to some reasons, failovers may fail with allocation issues. Therefore, it is recommended to identify the workloads that require high SLA for capacity availability in the DR location, based on business criticality or compliance requirements. For these critical workloads, you are recommended to opt-in for On-Demand Capacity Reservations when using Azure Site Recovery. When you configure capacity reservation group with ASR, Capacity Reservation SLA gets added for failovers in DR location. Please keep in mind that capacity reservation has cost implications. 

Screenshot of an Azure VM Disaster Recovery Section

How can cost be optimized for capacity reservations?

Capacity Reservations are priced at the same rate as the underlying VM size. To optimize cost when using Capacity Reservation along with ASR, you can also use Azure VM Reserved Instances or Azure Savings plan for compute. This is because capacity reservations are eligible for Savings Plan and Reserved Instances term commitments discounts.

Azure Reserved Instances or Azure Savings Plan do not provide any capacity SLA but provide discounted rates for commitments whereas Capacity Reservation provides the capacity SLA.  Keep in mind that reserved instances or savings plan only cover the compute cost of VMs; additional costs (like OS license cost, software licenses, etc.) will still apply. 

• Azure Reserved Instances allows you to commit to one-year or three-year term of compute capacity by paying up-front or monthly at a discounted rate compared to pay-as-you-go rates, enabling customers to save up to 72% for the VM cost.
• Azure Savings Plan allows you to reduce compute usage costs by up to 65% compared to pay-as-you-go rates by making an hourly spend commitment across all in-scope instances.

Depending on your workloads pattern and needs, you can decide between a savings plan and a reservation. If you have workloads that run continuously and are stable, Azure Reserved Instance may be the preferable option. If you have workloads that are dynamic and evolving, Azure Savings Plan may be a preferable option.

If you are using Azure Reserved Instances for the ASR protected workload configured for regional protection, note that the targeted region would not be covered by the Reserved Instance pricing that was being applied in the source region.  The target region would require its own Azure Reserved Instance commitment to cover the additional VMs in the failed over region.  Since failed over workloads are typically short term and tend to be failed back to the original source, you will need to determine if additional Reserved Instance commitments make sense for your situation.

If you are using Azure Savings plan, ensure ASR failed over VMs are within scope of the savings plan. With Azure savings plan, hourly usage charges incurred from savings plan-eligible resources, which are within the benefit scope of the savings plan, are discounted and applied to your hourly commitment until the hourly commitment is reached.  

Illustration  

Let us take an example of an Azure Reserved Instance:

• Source location of VMs: East US2
• Target location: West US2
• 2 VMs: each of D4s_v5 (4 vCPU, 16 GiB RAM), one running Windows utilizing Azure Hybrid Benefit and the other Linux. Both these VMs have ASR configured with the source VM size configured for ASR target as well. 
• Target resource group named: asr-westus2-rg

You have also associated a Capacity Reservation Group while configuring ASR on each of above VMs. This associated capacity reservation group has two (2) VM instances of size D4s_v5. By default, capacity reservation will utilize the pay-as-you-go pricing of the D4s_v5 VM.  The cost for this capacity reservation will be for two D4s_v5 VMs in West US2 per month (2*140.16 = USD 280.32). So irrespective of whether you have failed over to West US2 or not, you will keep paying USD 280.32 per month for capacity reservation.

Now, you purchase reserved instance for 2 VMs of size D4s_v5 in West US2 for a commitment term of 3 years with the scope as single resource group (asr-westus2-rg) which is same resource group used as target resource group for ASR. The price for reserved instance for 3-year term is (2*51.8592 = 103.7184) USD 103.7184 per month, thus having ~63% of cost savings. This reserved instance applies to your capacity reservation, and you are paying USD ~104 per month effectively.

Please note that prices in the illustration are from Pricing - Windows Virtual Machines | Microsoft Azure as on 08-Nov-2024. For Windows, Azure Hybrid Benefits (AHB) were also applied while making the above calculations. Prices are estimates only and are not intended as actual price quotes. Actual prices may vary.

Screenshot of price from Azure Pricing Page for Windows OS with AHB applied

Screenshot of price from Azure pricing page for Linux

Conclusion

Using Capacity Reservation along with Reserved Instances or Savings plan together can provide high SLAs for capacity availability at discounted rates. One thing to note before we conclude this article is that you can use Azure Site Recovery to perform global disaster recovery to replicate VM between any two regions if Azure Site Recovery supports them. These 2 regions can be non-paired regions as well.

Please note: Scope of this article is disaster recovery for Azure VMs.