Google continues to build out its cloud computing platform and today the company announced that its autoscaling service for Compute Engine, its infrastructure-as-a-service platform, is now available in beta.
Using this new feature, developers can now have Compute Engine automatically spin up new machines based on demand. If your CPU utilization goes above a certain value or your HTTP load balancer notices a spike in incoming traffic, for example, you can now have Google start a new machine to distribute that load. You can also connect the autoscaler to Google’s Cloud Monitoring API to select custom metrics that’s important for your application. This also means you don’t need to have machines on standby to take care of unexpected demand. Instead, they only spin up if needed, which could save you quite a bit of money.
The company previously announced that it would soon launch this feature at its Cloud Live event earlier this month, but it was never quite clear when it would become available. At the event Google showed how you could use autoscaling to go from a system that was barely handling and load to dealing with 1.5 million requests per second.
It’s surprising that it took Google this long to release this feature. Amazon made auto scaling available on AWS back in 2009 (unlike Google, Amazon prefers “auto scaling” over “autoscaling”) and Microsoft launched Azure auto-scaling (it prefers to use a hyphen) for web sites, cloud services and virtual machines last June. Previously, Google recommended that developers use a service on its App Engine service to automatically scale and orchestrate Compute Engine applications. That worked, but it was quite a bit more intricate and error prone than setting up a few metrics in the new autoscaler.