Introducing Heat Migrations on Akamai

As a part of our continued backend enhancements, we’ve developed a quicker technique to resize digital machines on our platform and considerably lower downtime throughout routine upkeep when shifting a VM to a special host. This was achieved partly by means of our purpose to design stay and chilly migrations to make use of the identical job sort. 

All through the lifetime of your server, you could have to resize your VM, change your plan sort, and even transfer to a different Akamai knowledge heart. Periodically, Akamai might also want to maneuver VMs to rebalance hosts or carry out routine upkeep.

Earlier than as we speak, there have been two forms of migrations on Akamai’s compute platform:

• Reside Migrations – With a VM powered on, we will transfer it to the same host throughout the identical knowledge heart with no down time.
• Chilly Migrations – With a VM powered off, we will resize it or transfer it to a special knowledge heart. This typically incurs vital downtime, particularly for bigger VMs.

We added a 3rd migration sort: heat migrations. Heat migrations permit VMs to stay powered on by means of most of a transfer, decreasing the downtime to lower than one minute. Like most cloud suppliers, Akamai runs a combined fleet atmosphere, and we now have an answer to relocate VMs shortly to completely different hosts. This may even allow a quick resizing or altering from shared or devoted plans to premium cases, which assure a particular processor mannequin or newer.

Heat migrations got here from the method of streamlining and growing the reliability of chilly migrations, which is one other replace we will proudly announce as we speak. All of those make up our unified migration platform. 

Reside and chilly migrations now use the identical job sort and run as a single job reasonably than a number of unbiased jobs that may succeed or fail. This removes any potential errors attributable to a VM current on two techniques after a migration job completes and the method is not going to modify the state of the document or the supply host till the disk and state have been efficiently synced to the vacation spot host. This considerably decreases the potential want for handbook intervention from our help crew when shifting a VM to a different area.

In abstract, we’ve much less overhead to keep up and could have a neater time with upkeep and rebalancing duties. For you, this implies:

• Heat migrations can transfer a VM to a different host in the identical knowledge heart in underneath one minute.
• Considerably much less downtime when resizing a VM or altering your plan.
• Considerably much less downtime once we rebalance or improve our fleet.
• Significantly elevated reliably when migrating from one knowledge heart to a different.

Sources