A master's thesis conducted in collaboration with Red Hat explores potential optimizations for Open vSwitch (OVS) by employing coflows to anticipate imminent network traffic. This approach aims to reduce the latency-inducing upcalls to the control plane, which are typically triggered by cache misses in the OVS megaflow cache. The study includes a series of benchmarks performed on an OVN-simulated, single-node OCP cluster. These benchmarks utilize XDP to timestamp packets at both the ingress and egress points of the cluster, measuring latency across various traffic scenarios. These scenarios are generated using synthetic coflow traffic traces with varying flow size distributions. The findings provide a comprehensive analysis of how OVS performance is influenced by accurately predicting different proportions of future flows under varying traffic conditions.

This paper presents a condensed version of the research, focusing on the key findings and methodologies used to improve Open vSwitch performance through coflow-aware branch prediction.