CPU in Kubernetes from the kernel up9/12NUMA, SMT and cache: when hardware topology starts to matter
Two workloads can get the same CPU time and do different amounts of work. This is where the CPU abstraction ends and the physics of the machine begins.