blog single gear

Cloud Foundry App Autoscaler: An Update

We wanted to share a quick update on the progress of the application autoscaler project, as we’ve recently hit some nice milestones.

What is this?

The app autoscaler is a Cloud Foundry extension (part of the Extensions PMC) that automatically scales Cloud Foundry applications based on load.

Why would I want it?

Do you like paying for more instances than you need? Do you like not having enough instances when load unexpectedly increases? If you, like me, want as many instances as you need and no more then you, like me, will like App Autoscaler.

Cool, who uses this already?

The big news – and the reason I’m writing this – is that I’m excited to announce the first GA production deployment of the App Autoscaler: the App Autoscaler is now used in production on the SAP Cloud. App Autoscaler’s design was initially based on the BlueMix App Autoscaler project which has been running in production and at scale for quite a while (on BlueMix, unsurprisingly). We’re hoping to be able to match all the features of the BlueMix autoscaler using the open source autoscaler component in the next couple of months at which point we will deploy the open source autoscaler to BlueMix also.

Is Autoscaler ready for me to use on my cloud?

I would cautiously say yes. The team feels that Autoscaler is ready to go. We want to wait until App Autoscaler is running in two production environments before officially declaring a 1.0 version, but the current version is stable and is working well at scale.

Who is contributing to the App Autoscaler effort?

I’m proud to say that App Autoscaler has been a truly cross-foundation project, and we’ve had committers from IBM, SAP and Fujitsu. We work using a distributed commit model (and when we say distributed we mean distributed — at various points we’ve had team members based in Australia, India, Seattle, China and the UK!).


Ok! You can get started with the Bosh release at We’re working on enabling deploying as a Cloud Foundry app (and removing the consul dependency) as an easier deployment model for some environments. Kick the tires, give us feedback (we’re on slack in #autoscaler) and – it goes without saying – contributions welcome!

Anything else?

On behalf of the Autoscaler team, thanks!

Julian Friedman Profile Image

Julian Friedman, AUTHOR