A Practical Guide to NUMA Affinity in Kubernetes

A Practical Guide to NUMA Affinity in Kubernetes

NUMA effects are one of those problems that don’t show up in dashboards, but will happily show up in your p99 latency and in “why is this box slower than the identical box next to it?” Kubernetes can help—but only if you enable the right node-level managers and verify the result from inside the workload. What is NUMA? Modern multi-socket servers split memory banks across CPU sockets. Each socket and its directly attached memory form a NUMA node (Non-Uniform Memory Access). Accessing memory on your own socket is fast (local); crossing the interconnect to another socket’s memory is slower (remote). ...

A Practical Guide to CPU Pinning in Kubernetes

A Practical Guide to CPU Pinning in Kubernetes

I recently attended the Cloud Native Days France 2026 conference in Paris, where Ricardo Rocha (Head of Infrastructure at CERN) opened the session with a talk titled “10 PB/s without breaking the budget.” He explained in detail how the LHC upgrade forced them to process 10 petabytes of data per second with a fixed computing budget. The solution was not magical hardware, but rigorous resource isolation. If strict resource limits are effective enough for particle physics, they are certainly effective enough for our production workloads. ...