Every week we share our expert TIP with you.
#35: Consider using Spot Virtual Machines
If reducing the costs of Virtual Machines in the cloud is one of your priorities, you may want to have a look at Spot Virtual Machines.
Most organisations have some workloads that are business-critical, while other workloads are significantly less important. It is possible to reduce the costs of those less important workloads by taking advantage of “Spot” Virtual Machines. Amazon AWS and Google have been providing Spot Virtual Machines (respectively called “Spot Instances” and “Preemptible Virtual Machines”) for several years now, and recently, Microsoft announced the preview of Azure Spot Virtual Machines.
What are these Spot Virtual Machines?
Spot Virtual Machines provide access to unused resources within a Cloud Provider at a significant discount. Savings of up to 90% can be achieved by using the Cloud Provider’s otherwise unused servers. The downside of using these Spot Virtual Machines however is that these resources can be recalled by the Cloud Provider almost immediately if the resources are needed elsewhere. Amazon AWS should provide a 2 minutes warning before terminating the Spot instance. From Microsoft, you will get a 30 second warning prior to a Spot VM being “evicted”.
When would using Spot Virtual Machines be a good choice?
There are several scenarios you could think of where running your workloads on Spot Virtual Machines would be ideally suited, such as:
- Batch jobs;
- Workloads that can sustain and/or recover from interruptions;
- Development and testing environments;
- Short-lived jobs which can easily be run again if the Virtual machine is stopped.
So generally, any workload where a VM suddenly being stopped by the Cloud Provider would not cause a huge problem. Either a workload that can continue with reduced resources and/or a workload that can simply pick up where it left off at a later moment.