Netflix Engineers Tackle Container Scaling Challenges on Modern Multi-Core CPUs

Summary

Netflix engineering has published insights into their infrastructure challenges with container scaling on modern high-core-count CPUs, a problem they've dubbed 'Mount Mayhem.' The issue centers on how container orchestration systems handle filesystem mounts and resource allocation across processors with dozens or hundreds of cores, which has become increasingly problematic as CPU architectures have evolved. Netflix's container infrastructure must efficiently manage thousands of containers across their global content delivery and streaming platform, making CPU resource management critical to performance and cost optimization.

The technical challenge emerges from the intersection of Linux kernel behavior, container runtime mechanics, and modern CPU architectures featuring numerous cores. As CPUs have scaled from 8-16 cores to 64+ cores per socket, certain operations that were previously negligible have become significant bottlenecks. Netflix's engineering team has identified specific pain points in how container mount operations interact with kernel scheduling and CPU affinity on these high-core-count systems.

This work reflects Netflix's ongoing investment in infrastructure optimization to support their massive streaming workload. The company serves hundreds of millions of subscribers globally, requiring sophisticated container orchestration to handle encoding, transcoding, recommendation systems, and content delivery. By addressing these low-level infrastructure challenges, Netflix continues to push the boundaries of large-scale cloud-native deployments and contributes valuable insights back to the broader infrastructure community.

• The research contributes to broader understanding of container orchestration performance at massive scale