A cross-layer reliability metric for wireless mesh networks with selfish participants

Abstract

Wireless Mesh Networks (WMNs) have emerged as the next generation networking technology that fosters the development of new network paradigms such as Wireless Mesh Community Networks (WMCNs) [1]. Many of the applications envisioned to run on WMCNs have high-throughput requirements. Several routing metrics have been proposed in recent years to select the path with the highest delivery rate in wireless multi hop networks in order to improve the throughput experienced by network terminals. The essence of all these metrics lies in the necessity of avoiding the selection of unreliable network paths due to the presence of lossy wireless links that are prone to transmission errors. However, in the presence of selfish mesh routers that drop the packets sent by other network nodes, these metrics fail to select the network path with the highest delivery rate and thus with the highest end to end throughput. Specifically, even the presence of only one selfish mesh router that drops almost all traffic on a path composed of highly reliable wireless links can lead to serious unfairness and throughput degradation. In this paper we propose a new cross-layer metric that combines information across routing and MAC layers to cope with the problem of selfish behavior (i.e. packet dropping) of mesh routers in a WMN and select the path with the highest packet delivery rate considering both the quality of the wireless links and the reliability of the nodes. Our metric, EFW, combines routing-layer direct observation of forwarding behavior of neighbors, with MAC-layer quality of the wireless links in order to allow a routing protocol to select the most reliable and high-performance path. We integrated the proposed metric with a well-known routing protocol for mesh networks, OLSR [2], and evaluated it using the NS2 simulator using the physical layer extension from [3]. Our preliminary results show that the proposed metric improves the network performance with respect to the baseline approach more than 200% when several selfish mesh routers are placed inside the network. Ongoing work consists of