🔴 Live show on Thursday: live migrations finally come to pods
- Sign up before the event to get a chance to win prizes.
- Add the event stream to your calendar.
- See you Thursday on YouTube Live.
For years (I count 3-7), we've seen researchers and open source engineers working on Linux tools to "live migrate" a container workload between nodes, but it never quite got there.
This idea of moving the storage, the in-memory data blocks, the pod network endpoint, and the IP connections has been a feature of virtual machines for over 20 years! (I first used it in ESXi and Windows Hyper-V around 2005-2009 and thought it was magic.)
But VMs have a clear demarcation line. The hypervisor knows exactly what storage, memory, and networking belong to a specific VM. Now, within a specific OS, it wasn't so obvious to track app dependencies of a Linux binary until Docker and Kubernetes came along. This gave us the potential for using a set of APIs to track all those resources and send them from one Linux Kernel to another across the network.
Join us Thursday as my guests from Cast AI help us understand how this milestone is achieved (hint, it involves CRIU).