Stackato platform revision auto-scales app instances, this time for real
One of the benefits of deploying an application over a cloud PaaS service like ActiveState's Stackato is to scale the performance of that application as user demand over the network requires. But this often makes it incumbent upon the developer (the Stackato customer) to know when such demand takes place, in order to respond by doing the necessary scaling.
Is there any way to automate this task? In a 2013 webinar--an eon ago, in the cloud era--ActiveState technology evangelist John Wetherill said automating the scaling process would essentially be impossible. "Instead of us baking in a whole lot of complex rules and policies to decide when and how to scale an application--which is almost an impossible task for us to do, because every customer would have different requirements," Wetherill explained at the time, "we decided we would open up the API, allowing a user of Stackato to easily monitor the performance of the applications... and then call back into Stackato to scale."
Well, maybe the cloud industry evolves so quickly as to make the impossible possible in one year's time, but with this week's rollout of version 3.2 of the Stackato platform, ActiveState successfully implemented application auto-scaling. Customers no longer need to employ third-party monitoring tools, or write their own scripts and call it automation.
In the new edition of the Management Console, according to new documentation, a customer can activate the auto-scaling option by specifying the maximum CPU utilization rate which triggers instantiating a new instance in the application pool, and the minimum utilization rate which triggers spinning down an instance. Alternately, customers can specify this option within their configuration files.
The concept of application auto-scaling has been brought up in the context of other PaaS platforms in the past, though one must be careful to take note of what their practitioners mean by "auto." In 2011, RightScale CTO Thorsten von Eicken tackled the subject with respect to Cloud Foundry by explaining why CF did not autoscale, before envisioning how RightScale could resolve that issue by monitoring the platform's "slack pool" and spinning up new servers when the pool runs dry. This started a trend where autoscaling became feasible on any number of platforms, provided the right tools were in place and the right conditions were set.
This week, ActiveState is making clear that for Stackato 3.2, "auto" is not only short for "automatic" but "automated," stating that the process takes place completely within the Stackato platform without the use of third-party monitors or scripts.
- see the ActiveState Stackato 3.2 download page