DECON: Decentralized Coordination for Large-Scale Flow Monitoring

Monitoring at the flow level is crucial to ensure the correct operation of networks. Any sizable network relies on a number of monitoring probes, both to provide different observation points but also to scale to the ever-increasing number of flows that go through it. This situation gives rise to the difficult problem of assigning monitoring of flows to the available probes so that the network-wide coverage of flows (i.e., the number of flows actually monitored) is maximized. In this paper we introduce DECON, a decentralized and scalable coordination system aimed at solving this assignment problem. Unlike other approaches, DECON requires no network topology information, no traffic matrices and no packet marking. We present extensive simulation results showing that DECON scales up to large numbers of flows while requiring reasonable amounts of state from probes. Further, performance results from a prototypical monitoring probe built with commodity hardware show that even an inexpensive solution can accommodate DECON's requirements.