Google brings autoscaling to Compute Engine

Google has activated the so-called Autoscaling functionality on its Compute Engine platform. This mechanism can quickly switch virtual machines on and off based on various parameters, such as the number of requests. According to Google, this can lead to cost savings.

The Compute Engine Autoscaler detects the demand for, for example, a website on the basis of various parameters. For example, it can look at the CPU load of a group of available instances, virtual machines within the Compute Engine platform. The Autoscaler can also request the number of queries per second from http load balancers. This information allows the Autoscaler to activate extra instances when it is very busy, while the number of active instances can be dynamically reduced during off-peak hours.

According to Google, Autoscaler also chooses the optimal number of instances based on other parameters. For example, it is also checked from which location internet users request information. This mechanism could better absorb a sudden spike in traffic. Google claims that the Autoscaler is able to activate enough instances from a scenario with virtually no HTTP requests to meet a peak demand of 1.5 million queries per second.

According to the internet giant, the Autoscaler not only offers the Compute Engine user the opportunity to better cope with busy and quiet periods, it could also provide significant cost savings. The potential savings partly depend on the settings the user chooses in their management console.

The new functionality will be offered to customers by Google for the time being under a beta label. As a result, the use of the Autoscaler does not fall under any slas that are closed. Also, Google may make changes to the Autoscaler API that may not be compatible with the current implementation.