Optimization based queue management for opportunistic network coding

In existing opportunistic network coding architectures, such as COPE, due to asynchronous packet transmissions and rate mismatch, the amount of packets which can be coded together may be insufficient. This may limit the performance gain of network coding. Purposely delaying some packet transmissions to accumulate more coding opportunities could be a feasible solution to this problem. However, excessive delay may be harmful to delay sensitive applications, such as video streaming. In this paper, we propose to use queue management including packet delaying and scheduling mechanisms to maximize coding opportunity while meeting delay constraints of traffic flows. We formulate the design of delaying policy as an optimization problem. Based on the model, we propose a distributed network coding-aware queue management policy (DQM) where each node independently decides whether and for how long to delay a packet. We use simulation experiments based on ns2 to evaluate the performance of DQM. Numerical results validate our analytical model and show that DQM can substantially improve the performance gain of opportunistic network coding compared with COPE.