Boot faster! Faster!!

A while ago some complained about the time it took to bring up large amounts of instances in Scalr. While never a problem for scaling web apps (incremental scaling), it was for bringing up data crunching farms, crawlers, and any kind of asynchronous task.

We looked into what optimizations we could do to help you with that, and found quite a few.

The Scalarizr initialization and messaging phases have been reduced from a couple minutes per instance to less than 10 seconds for a whole farm. In some tests we ran, we had 50 instances send HostUp within 7 minutes of starting a farm with min_instances = 100, which includes API call, provisioning, and OS boot. For instances that need EBS volumes, boot and initialization take longer because of the AWS requirements, but you should still see about a 50% speed increase (33% time to HostUp decrease) in those cases.

I hope you enjoy this effort!

Cheers,
The Scalr “Make it faster! Faster!!!”  team