Evaluating CRIU with Pre-Copy Memory Migration

Stateful container live migration remains a critical bottleneck in container orchestration despite advancements in stateless migration techniques. While stateless workloads migrate seamlessly, stateful systems such as databases, message queues, and file systems face unique challenges in maintaining state persistence and consistency during transit.

This empirical study critically evaluates these challenges through production-level benchmarking of CRIU (Checkpoint/Restore in Userspace) with pre-copy memory migration. By testing diverse workloads, the research identifies precise failure thresholds in write-heavy databases and latency-sensitive microservices under real-world conditions.

The analysis reveals that current migration techniques exhibit brittle performance under production stress. Under write-heavy loads, memory dirtying rates increase exponentially, causing migration times to stall when modification rates exceed 10 Gbps network capacity. Furthermore, network switchover introduces latency spikes of up to 400ms, resulting in request timeouts in 15% of cases.

Stock photo for illustration only, not from the actual event