Container Checkpoint - Restore at Scale for Fast Pod Startup Time

Explore container checkpoint/restore techniques for faster pod startup times and improved cold start performance in this conference talk from KubeCon + CloudNativeCon Europe 2022. Discover how MathWorks cloud infrastructure achieved both rapid startup and pre-warming goals using Container Checkpoint/Restore. Learn about design considerations, lessons from years of production experience, and best practices for implementing container checkpoint/restore in Kubernetes without native support. Gain insights into improving system scalability and utilization, and explore a vision for native CRIU support in Kubernetes. Delve into topics such as scaling options, checkpoint/restore mechanics, cluster architecture, implementation steps, and future enhancements for optimizing container-based applications.

Intro
During this talk...
Our goal is to create a scalable system...
But achieving all three goals is challenging
Option 1: Scale out on-demand with usage
Option 2: Create pre-warmed standby pool
Is there a way to achieve all goals?
Yes, we can!!! Checkpoint/Restore
Checkpoint/Restore: Behind the Scenes
Eliminating Cold Starts with Checkpoint Restore ..
Kubernetes and Checkpoint/Restore
Our Approach of Non-Native support
Container Runtime in Container Runtime
Checkpoint sequence
Restore Sequence
Cluster Architecture with CR
Zero to Checkpoint Restore in Kubernetes
Lessons Learned
Best Practices
Future Enhancements
MathWorks is hiring....