Demand-aware centralized traffic scheduling in Wireless LANs

Abstract

A heavy deployment of IEEE 802.11 Wireless LANs and limited number of orthogonal channels make lots of Access Points (APs) overlap their interference regions, which greatly increases interferences between APs and stations. In order to cope with the performance degradation caused by the interferences, we propose Co-Fi, a centralized Wi-Fi architecture that effectively coordinates downlink transmissions by APs and improves network performance in terms of throughput and end-to-end delay. Co-Fi adaptively allocates time slots for APs and stations based on both traffic demands on the stations and a conflict graph that represents interference relationships among the devices. The scheme allows APs in exposed node relationship to use the channel simultaneously by setting the same backoff time. It also effectively avoids downlink conflicts created by hidden node and non-hidden/non-exposed node, by allocating non-overlapping time slots to interfering stations. To implement these adaptive traffic schedules, we design CoMAC, a hybrid MAC protocol at APs. Our evaluation results show that when APs are densely deployed and the network is highly loaded, the scheme achieves 3–5 times more throughput gain than Centaur, a state-of-the-art scheme while its end-to-end delays are 10–90% lower than those of Centaur and CSMA/CA.