FlowME: Lattice-based Traffic Measurement

Abstract: Flow-based traffic measurement is a very challenging problem: Managing counters for each individual traffic flow in hardware resources knowingly struggle to scale with high-speed links. In this paper we propose a novel lattice theory-based approach that improves flow-based measurement performances and scales by keeping the number of the maintained hardware counters to a minimum (result mathematically established in the paper). The crucial contribution of the lattice is to map the computational semantics of the packet processing to user requests for traffic measurement thus allowing for a better-informed and focused counter assignment. An implementation over an Openflow switch, FlowME, was developed and evaluated upon its memory usage, performance overhead, and processing effort to generate the minimal solution. Experimental results indicate a significant decrease in resource consumption.