Approaching Optimal Centralized Scheduling With CSMA-Based Random Access Over Fading Channels

Carrier sense multiple access (CSMA) based distributed algorithms can attain the largest capacity region as the centralized Max-Weight policy does. Despite their capability of achieving throughput optimality, these algorithms can either incur large delay and have large complexity or only operate over nonfading channels. In this letter, by assuming arbitrary backoff time, we first propose a fully distributed randomized algorithm whose performance can be pushed to the performance of the centralized Max-Weight policy not only in terms of throughput but also in terms of delay for completely connected interference networks with fading channels. Then, inspired by the proposed algorithm, we introduce an implementable distributed algorithm for practical networks with a reservation scheme. We show that the proposed practical algorithm can still achieve the performance of the centralized Max-Weight policy.