P5: Policy-driven optimization of P4 pipeline

摘要: The physical pipeline of flexible network switches is usually programmed using packet-level programs, such as P4 programs. However, those programs are low level and leave room for further optimization. We propose P5 (Policy-driven optimization Pipeline), a system that exploits knowledge application deployments embedded in high-level policy abstraction to: 1) detect features used by applications mutually-exclusive way thereby remove inter-feature dependencies between the tables implementing these switch. This improves concurrency hence its efficiency. 2) not any application/traffic on switch given topology. reduces number resource consumed switches, which also Our experiments real show resulting pipelines up to 50% more efficient compared cases do exploit this information.