This article is more than 1 year old

Microsoft previews Azure App Service Automatic Scaling, for when defining your own rules is too much like hard work

What could possibly go wrong?

Microsoft has rolled out an early preview of Azure App Service Automatic Scaling – a handy tool, assuming Azure hasn't hit capacity once again.

The Automatic Scaling feature is designed to do away with all the pesky rules and schedules of Azure's existing Autoscale feature in favour of something managed by the platform itself.

When enabled through either the Azure CLI or ARM templates, the function will automatically scale out the number of running app instances as the request count increases and scale things down again when demand subsides. Developers can slap limits and minimums on the service; the latter being useful for dealing with bottlenecks, while the former is essential to avoid a heart stopping invoice.

Once configured, Automatic Scaling takes precedence over existing Autoscale rules and schedules. Microsoft also recommends disabling health checks for web apps where the feature has been turned on.

"The health check requests can cause unnecessary fluctuations in HTTP traffic," the company explained. A shame because health check is a useful way to see if one's app is having an unscheduled lie-down.

A cynic might suggest Microsoft could use such a tool for its own infamously wobbly efforts in the cloud.

As for cost, billing is by the second and the preview works with existing Premium Pv2 and Pv3 SKUs. It is also only for the Azure App Service for Windows and Linux and, astonishingly, Microsoft has thus far resisted the urge to tinker with the Azure Portal UI; configuring the preview must be done via Azure CLI or ARM template.

While Autoscale has long existed in the Azure world, defining the metrics and load schedules can be a pain, although having full control over how and when scaling happens will doubtless remain a preference for cautious administrators.

Microsoft's cloud rival, AWS, also features a variety of scaling options, including the vaguely creepy-sounding Predictive Scaling, which uses machine learning models to predict a customer's traffic and resource requirements. ®

Similar topics


Send us news

Other stories you might like