Coding Network Congestion Control And Qos

It is confirmed theoretically that network coding can achieve multicast capacity. The successful application of network coding in P2P network also reveals its advantage in terms of increasing network throughput. Network coding merges content of different data packets together so that every packet has the same significance to the receiver. In other words, coded packet is less sensitive to loss, which indicates its possibility to improve transmission reliability. Although network coding owns the above virtues, it is not massively employed in real network. One of the underlying reasons is the absence of corresponding congestion control and QoS mechanism, which is exactly the objective of this thesis.The classical tool used to design congestion control algorithm is the feedback control system derived from network utility maximization (NUM) problem, so we first formulate a class of NUM problem for network with coded multicasts. Based on the primal and primal-dual form of the problem, we propose two congestion control models for network coding, whose stability is demonstrated through Lyapunov criteria. Although a few researches have discussed the optimization problem under network coding scenario, their aim is to seek minimum-cost multicast sub-graph. In contrast, this paper directly associates utility optimization with congestion control. The derived algorithms can handle network dynamism with only end-to-end information exchange, and have the form of TCP-like congestion control.Since transmission delay usually cannot be ignored in real network, we further derive the sufficient condition for local asymptotic stability of congestion control model with delay. The result provides suggestions on parameter configuration of congestion control algorithm for coded flow. A complete congestion control strategy has to include active queue management (AQM) mechanism at intermediate node as well. Hence, from the standpoint of optimal control, we establish a mathematical framework for AQM design under network coding.To support QoS in network with coded flows, we introduce a packet scheduling and buffer management algorithm, which runs at router and can guarantee QoS. The new algorithm removes the impractical hypothesis in previous researches that the size of buffer storing coded packets is infinite, and calculates packet delivery and drop rate on links through optimization problem, so it can also be regarded as an AQM mechanism for network with coded packets.