Get More Computing Power by Helping the OS Scheduler
Discover how to maximize CPU performance in Kubernetes workloads through improved process scheduling and memory locality. Learn about the CRI Resource Manager (CRI-RM) and its ability to dynamically manage CPU and memory pinning for all Kubernetes containers on a node. Explore case studies demonstrating significant improvements in database and AI training performances without workload-specific configurations or changes to upstream Kubernetes components. Gain insights into the problem of thread scheduling on CPU cores, the features of CRI-RM, its operational mechanisms, and future developments in this 20-minute conference talk presented by Intel experts Antti Kervinen and Alexander Kanevskiy at a Cloud Native Computing Foundation (CNCF) event.