Today, we’re introducing a feature to Google Compute Engine Autoscaler and Managed Instance Groups that enables region-specific control when scaling compute resources. This is important for customers that need high availability.

Called Regional Instance Groups, this new feature lets you set-up infrastructure that automatically spreads VM instances across three zones of the region and keeps the spread equal as you change group size. Autoscaler works regionally across all three zones, adding and removing instances so that the spread remains equal.

Back in September, we announced the general availability of Google Compute Engine Autoscaler and Managed Instance Groups. Since then, we've seen amazing growth in usage and heard lots of feedback. One of the top requests from customers was to simplify configurations allowing for higher availability.

Now, with region-specific control, in the rare event of a bad build, network issues or zone failure, you're better protected against losing your service. Since your VM instances are equally spread across three zones, two-thirds of them will not be impacted. On top of that, when you’re using Autoscaler, it will notice increased traffic in other zones and adjust their number accordingly. After the faulty zone goes back to normal, the load will be rebalanced to use all three zones again.

Setup is easy. If you choose to use regional configuration, simply select Multi-Zone group while creating a Managed Instance Group. At this point, your choice of locations are regions and the scope of your Managed Instance Group and Autoscaler becomes regional.
To use Regional Instance Groups in Alpha and learn more about the service sign-up here.

- Posted by Jerzy Foryciarz, Product Manager, Google Cloud Platform