Introduction

Lightwing ComputeGroups is a cluster management product. It lets you run high availability clusters composed of Spot and Ondemand instances. Spot instances typically comes with a 70% cheaper rate than the ondemand instances. They are pretty good to run variety of workloads.

Cloud providers usually don’t provide any availability guarantee for spot instances. Because of this restriction, it becomes challenging to run production workloads on Spot instances reliably. Lightwing ComputeGroups addresses this issue. With ComputeGroups, you can run long running spot instances and the entire spot workflow is handled by Lightwing.

ComputeGroups is not replacing any of the native ways to run a workload on the cloud. It integrates with cluster management products like AWS Auto Scaling groups, Azure Scale sets etc and add additional capabilities to it.

To understand how ComputeGroups differs from native cloud capabilities, read this blog on the subject.

Benefits

• 70-80%+ cost savings
• Hassle free automation and management of cloud excess capacity spot instances
• High available, cost optimized cluster of instances

How it works

You will configure ComputeGroups with the required capacity settings, instance types and availability zones. ComputeGroups will launch instances matching to the specified configuration. ComputeGroups ensures that most of the instances in the group are spot instances and they are selected based on the least interruption possibility.

ComputeGroups also maintains a historical data about Spot instances and interruptions. This helps ComputeGroups to choose the right set of instances for the given workload. ComputeGroup handles spot interruption notices and perform graceful draining of the workloads.

Application’s business criticality and other metadata can be configured on ComputeGroups which lets users control the spot instance placement to maximize savings and increase graceful draining.

IaC and Multi-cloud first

ComputeGroups configuration works transparently across multiple cloud providers. This makes moving workloads between multiple cloud providers easy and seamless.

It can be configured with modern Infrastructure as Code tools like Terraform, Cloud Formation etc.