Throughput-delay-reliability tradeoffs in multihop networks with random access

In ad hoc networks, performance objectives are often in contention with each other. Indeed, due to the transmission errors incurred over wireless channels, it is difficult to achieve a high rate of transmission in conjunction with reliable delivery of data and low latency. In order to obtain favorable throughput and delay performances, the system may choose to compromise on its reliability and have nodes forcibly dropping a small fraction of packets. The focus of this paper is on the characterization of tradeoffs between the achievable throughput, end-to-end delay and reliability in wireless networks with random access. We consider a multihop ad hoc network comprising several source-destination pairs communicating wirelessly via the slotted ALOHA channel access scheme. Employing ideas from statistical mechanics, we present an analytical framework for evaluating the throughput, end-to-end delay and reliability performances of the system. The main findings of this paper are (a) when the system is noise-limited, dropping a small fraction of packets in the network leads to a smaller end-to-end delay though the throughput suffers as well, and (b) when the system is interference-limited, however, there exist regimes where dropping a few packets in the network may actually reduce the end-to-end delay as well as increase the system throughput. We also present some empirical results which corroborate the results obtained analytically.