How It Works

SpotSurfer Checkpointing and Recovery Service

The cloud, especially a Spot instance, is cost-effective for fault-tolerant, cloud native apps. Not necessarily so for thousands of stateful, long-running, non-fault-tolerant apps.

While spot instances can save up to 90%, there is one big catch.  The cloud service provider can take them away any time they need the capacity.  Customers get a notification ranging from only 30 seconds to 2 minutes before the instances disappear.  The result is spot instances are not conducive to applications that cannot handle this kind of infrastructure volatility.  Imagine a long-running app, almost getting to the finish line, but getting the spot eviction notification and having to re-start from the beginning. That’s why cloud service providers recommend running only stateless or fault-tolerant apps on spot instances.

SpotSurfer is an application continuity service that uses the AppCapsule checkpointing and restore service + cloud automation to allow stateful workloads to run safely on Spot instances, saving up to 90% in compute cost.

It’s simple. Register your app container in the Memory Machine Operations Center, submit your job and the SpotSurfer service takes it from there. After receiving a spot eviction notice from the Cloud, Memory Machine will generate an AppCapsule. It will then coordinate with the cloud service to allocate a new spot instance, restore the AppCapsule on the new instance and the workload will continue.  The entire process is transparent to the application and fully automated.

How SpotSurfer Works

SpotSurfer Case Study