Delay gains from network coding in wireless networks

Abstract

We consider a single-hop cellular wireless system with a single source (base station) broadcasting a stream of incoming files to multiple receivers over stochastic time-varying channels with non-zero erasure probabilities. The base station charges a price per receiver per file with the aim of maximizing its profit. Customers who wish to transmit files to the receivers decide to enter the system based on the price, the queuing delay, and the utility derived from the transaction. We look at network coding and scheduling as possible strategies for file transmission, and obtain approximate characterizations of the optimal customer admission rate, optimal price and the optimal base-station profit as functions of the first and second moments of the service time processes under mild assumptions. We show that network coding leads to significant gains in the base station profits as compared to scheduling, and also demonstrate that the optimal network coding window size is highly insensitive to the number of receivers, which suggests that pricing and coding decisions can be decoupled. We also investigate the behavior of network coding in the case where the number of receivers is sufficiently large, and derive scaling laws for the asymptotic gains from network coding. We subsequently propose a way to extend our analysis of single-source, multiple-receiver systems to multiple-source, multiple-receiver systems in general network topologies and obtain explicit characterizations of the file download completion time under network coding and scheduling, also taking into account the effects of collisions and interference among concurrent packet transmissions by two or more sources.

(cont.) Our formulation allows us to model multi-hop networks as a series of single-hop multiple-source, multiple-receiver systems, which provides a great deal of insight into the workings of larger and denser multi-hop networks such as overlay networks and peer-to-peer systems, and appears to be a promising application of network coding in such networks in the future.