End-to-end performance aware association mechanism for wireless municipal mesh networks

Abstract

In a wireless municipal mesh (muni mesh) network, a client station (STA) needs to associate with a mesh access point (MAP) for network access. Previous association mechanisms assume high-speed backhaul and only the access link being the bottleneck. This assumption holds for most WLANs, but in wireless mesh networks traffic could be bottlenecked either by the access link or by the bandwidth-limited wireless backhaul. In this paper, we propose a new joint MAP association mechanism for wireless muni mesh networks to improve STA’s end-to-end communication performance. A STA makes its association decision by jointly considering the quality of the access link between the STA and the associated MAP as well as the cost of the multi-hop path from the associated MAP to the Internet gateway. In addition, we design two new metrics, Contention Aware Expected Transmission Time (CAETT) and Load Aware Expected Transmission Time (LAETT), to measure the access link quality. The main strength of CAETT is incorporating the impact of 802.11 MAC layer contentions on the bandwidth sharing of multi-rate stations. LAETT further captures the real traffic load on the shared medium. In order to reduce the association delay, we use an analytical model to derive a hybrid measurement/estimation method to enable a station to quickly determine the cost of CAETT and LAETT. We conduct extensive simulations to evaluate the performance of the proposed joint MAP association mechanism. Especially within the joint association framework, we investigate the impact of various combinations of access link metrics (RSSI, PB, CAETT, LAETT) and backhaul routing metrics (hopcount, ETT, RALA) on the system performance. We show that the joint association mechanism can significantly improve the network performance in terms of throughput and delay by up to 100%. In particular, the joint association mechanism with LAETT as the access link metric and RALA as the routing metric outperforms other schemes and metrics.