Optimization of Network Flow Monitoring

A flow cache is a fundamental building block for flowbased traffic processing. Its efficiency is critical for the overall performance of a number of networked devices and systems. The efficiency is mainly dependent on a utilized replacement policy. This work proposes an approach based on Genetic Algorithm. The proposed approach starts from recorded traffic traces and uses Genetic Algorithm to evolve innovative replacement policies tailored for the flow cache management in particular deployments. An extension of the replacement policy is proposed to improve the already optimized policy even further. The extension is based on an evolution of a replacement policy and a classifier of packet-header fields. The results show a consistent decrease in an eviction ratio in case of two considered problems – reduction of overall number of evictions and reduction of eviction in case of heavy-hitting flows.