Cloud Reserved Instances: Good Discounts for Limited Use Cases
Find out how the reserved instances pricing model for cloud services works and in what scenarios it makes sense.
April 3, 2018
More and more companies are moving their infrastructure and applications into the cloud (rightly so, I’d say) and while they are planning for this move, a very important decision to be made is how to execute the move: Will servers be lifted and shifted? Are they going to be rebuilt on the new platform or are they going to be completely transformed into more “cloud-native” applications using platform as a sService (PaaS) offerings, for example.
For most companies the answer most likely is going to be “all of the above depending on the application.”
With that decision comes the hard reality about lifting and shifting VMs to a cloud provider: VMs are expensive, especially if they run 24/7 every day of the year. (For many reasons, it doesn’t make sense to shut VMs down or scale them in/out depending on load, but I will not delve into that issue here.)
To make it attractive for customers to move those workloads to the cloud, all major cloud providers now offer “reserved instances” to customers. The way it works: Customers commit to using a certain number of VMs over a (usually one or three years) and in turn receive quite a decent rebate on the compute cost.
As the latest entrant to the reserved instances game, Microsoft offers some very flexible terms to Azure customers. (This article focuses primarily on Microsoft Azure reserved instances, but AWS offers EC2 reserved instances with discounts as high as 75 percent over on-demand pricing, according to the company, while Google Cloud’s “commited use” pricing discounts the normal monthly fee by as much as 57 percent.)
Significant Cost Savings
To demonstrate the cost savings on Azure, let’s look at the Australia Southeast region and a fairly common VM : a v2 Windows VM.
If this VM ran for one month, which in the Microsoft Azure universe is 730 hours (because 365 days times 24 hours divided by 12 months equals 730 hours), it would cost you roughly USD. Remember though, this covers only compute resources and the Windows license, not storage or anything else.
Now, if a company knew that the VM had to run for at least a year, it could contract for a reserved instance of that VM type in that region in that subscription and prepay the full yearly amount. In that scenario, the price drops significantly, by 41 percent.
Companies can save even more by committing to a three-year term upfront, or, in the case of Windows, by using the Hybrid Benefit model and bringing their own license.
When comparing the Microsoft Azure reserved instance model to the AWS EC2 reserved instance model, you’ll see that Microsoft allows customers to cancel their reserved instance early with a 12 percent cancellation fee, whereas AWS customers can’t get out of the arrangement unless they can sell their reserved instance on the AWS marketplace, which can be difficult.
Should I Use It?
Does this not sound great? All you need to do is lift and shift your VMs from on-premises to the cloud, never touch them again and the cloud providers will give you great discounts the longer you commit to running them.
Well, yes and no.
Cloud providers definitely offer a good incentive for IT shops to move VMs to the cloud and without a great deal of change to the workload receive some discounts for the first n months while the application owner and business decide what to do with the application (continue using the VMs as is, transform them or decommission them).
But I strongly advise that you do not view reserved instances as the blanket answer every time “What are we going to do with this VM?” is asked. It is a good tool to quickly onto a cloud and leverage some cloud technologies (logging, metrics, alerting, etc), but it should not be a long-term solution for all VMs, since there is often more value in focusing on transformation of applications so that they can leverage higher-level services like containers or even functions as a service (FaaS).
There will always (for the foreseeable future) be those pets around that are too expensive to be automated, make immutable, make auto-scalable or to transform, for various reasons, and for those applying a reserved instance discount might be a good approach. But using VMs in general should be the exception, not the rule.
Looking for an awesome, no-nonsense technical conference for IT pros, developers and DevOps? IT/Dev Connections kicks off in Dallas in 2018! And make sure to check out the cloud track, for which David O’Brien serves as the chairman, to hear more about moving workloads to the cloud, to find out how to modernize them and to learn from those who have already gone down this road.
About the Author
You May Also Like