Achieving proportional loss rate differentiation in a wireless network with a multi-state link

Many algorithms for providing the proportional loss rate differentiation (PLD) model have been proposed under a wired network. However, these algorithms suffer from low performance in a wireless network, which has a multi-state link, a link with location-dependent and time-varying channel capacity. This paper proposes two channel-aware and debt-aware droppers, named no-debt reference (NDR) and random probability (RP), to provide PLD in a wireless network with a multi-state link. The two proposed droppers prefer dropping the packet destined to a poor channel to improve the performance, causing some loss debt, which will be compensated later to keep PLD. From simulation results, NDR and RP actually achieve similar accurate loss rate proportion, lower queueing delay and loss rate, and higher throughput, compared with other methods in the wireless environment.