New Version of Elastigroup Autoscaler for Docker Swarm

Today, we are happy to announce the release of our latest version of Docker Swarm Autoscaler which includes new and improved functionality!

Previously, the Spotinst Docker Swarm Autoscaler was simply scaling clusters up to support unscheduled tasks that required more infra, or to ensure that pre-configured “headroom” units (for buffer) were available.

Scaling down was based purely on resource utilization metrics.

Shifting our focus to an application-first approach, the new version uses bin-packing algorithms which constantly run simulations of the cluster to ensure that only the minimal amount of instances required to support current tasks, are spun up (of course, the ability to add “headroom” units for buffer is still part of the Autoscaler). 

Likewise, for scale-down, the algorithms are simulating whether the cluster can support the tasks with optimized distribution, on less infra. If yes, scale-down will occur accordingly.

Additionally,  we’ve added support for placement constraints so you can control the nodes specific tasks are assigned to.

 Moreover, the new Autoscaler adds an extra security layer by implementing encrypted communication protocols with the Docker Master.

The new functionalities improve resource utilization and increase savings for Docker Swarm users even further.

More features are on their way, so stay tuned!