An Efficient Flow Label Collector Design for High-Speed Networks

Abstract

Flow label collection plays an important role in high-speed network traffic measurement. However, most existing measurement studies assume the flow labels are known a priori. Only a few methods are proposed to collect flow labels from network packet streams, which still have limitations like expensive communication cost and high missing rates. This letter addresses the limitations of the prior art by proposing an efficient flow label collector. It bounds the duplications of large flows and discards the small flows that do not matter, achieving lower missing rates at low communication cost. The experimental results based on real network traces show that, compared to the state-of-the-art methods, it can reduce about 90% flow label missing rate for medium flows or save more than 75% communication cost.