Flow-oriented network coding architecture for multihop wireless networks

Recently, a network coding-based packet forwarding architecture COPE was proposed for multihop wireless networks to improve the throughput. Both simulation and testbed results verified COPE's capability of enhancing throughput. However, COPE simply classifies all packets destined to the same nexthop into small-size or large-size virtual queues and then, to limit packet reordering, examines only the head packet of each queue to find coding solutions. Theoretically all the packets of distinct flows have the potential to be encoded together for throughput improvement, but the above packet size-oriented queue structure significantly limits this potential coding opportunity, since only one packet of a given size will be examined in the coding process. In this paper, we apply the standard flow-oriented queueing to network-coding-capable multihop wireless networks, which maintains a dedicated virtual queue for each flow and regards the head packets as coding candidates. Such flow-based architecture completely eliminates packet mis-order problem, and our theoretical analysis shows that in comparison with the simple size-based queuing, the new architecture is able to dramatically increase the potential coding opportunities. We further study the optimal network coding problem of the flow-oriented architecture. The COPE's simple coding algorithm is not effective enough since it does not consider several other key issues like the packet size gaps among the packets to be encoded, packet loss rates and decoding probabilities. Therefore, we develop a corresponding efficient algorithm for searching good coding solutions. Our extensive simulation results demonstrated that COPE can improve the node transmission efficiency, but the flow-oriented architecture with proposed coding algorithm can make this improvement much more significant. (C) 2011 Elsevier B.V. All rights reserved.