Open Networking Summit Europe 2019

Kubernetes is becoming the choice orchestrator for containerized deployments, with users flocking to it. More workloads are becoming performance sensitive to achieve the results now expected of them. K8s has started to evolve to cater for such workloads with the creation of the CPU and Device Manager to handle the compute and device resources on a K8s node. However, these components act independently of each other, which results in sub-optimal resource allocations. These less than ideal resource allocations can severely impact performance sensitive workloads. We are working in the K8s community to introduce a Topology Manager component that will co-ordinate these resource allocations, to ensure optimal performance.This talk will introduce and demonstrate the Topology Manager, showcase benchmark results showcasing its benefits and finally discuss the different use cases this applies to.