Scaling end-to-end measurements in heterogeneous wireless mesh networks

In large scale deployments of Wireless Mesh Networks (WMNs), access to the Internet is ensured by multiple gateways spread over the network. In such environments that rely on heterogeneous backhaul technologies offering different and time-varying bandwidth, delay or jitter characteristics, monitoring the end-to-end performances on the diversity of paths Internet flows can be forwarded on is challenging: the end-to-end measurement strategy must capture the diversity of backhaul connections, as well as multi-hop behavior within the mesh, and it must scale with the number of gateways, nodes and flows. In this paper we propose and evaluate the scalability of two measurement strategies for the monitoring of end-to-end paths. We establish closed form formulas for the overhead incurred by these measurement strategies, and compare their efficiency against greedy measurements in grid topologies. We conclude that one can reach linear increase in the number of probing nodes in place of an exponential growth for greedy end-to-end measurements. We extend these results to also show that this strategy takes advantage of dense topologies.