Characterizing multi-way interference in wireless mesh networks

Wireless mesh networks (WMNs) have been proposed as a solution for ubiquitous last-mile broadband access. A critical limiting factor for many WMN protocols in realizing their throughput potential is the interference between nodes in the WMN. Understanding and characterizing such interference is important for a variety of purposes such as channel assignment, route selection, and fair scheduling. Instead of using ad hoc heuristics, a recent study proposed characterizing interference in a WMN by measuring two-way interference, i.e., interference between each pair of communicating links.In this paper, we study the extent of multi-way interference, i.e., the interference caused by multiple transmitters to a communicating link. We find through simulations and through measurements of a 32-node wireless testbed that even if these transmitters individually do not interfere significantly with a given communicating link, simultaneous transmissions of them have the potential to significantly affect the throughput of the communicating link. This implies that pairwise interference measurements may be optimistic when used to drive protocols in wireless mesh networks. Encouragingly, we find that this phenomenon, although significant when it occurs, is not widespread. In particular, multi-way interference caused significant additional throughput degradation compared to pairwise interference to a small fraction of the links in the testbed over our measurement period. In addition, we find that there is a strong correlation between the impact of multi-way interference and the quality of the link under consideration. We conclude with recommendations on how protocols should take multi-way interference into account.