The Study Of Sliding Window Based Random Linerar Network Coding

This paper first described the research of network coding, and a review was given about the calssfication and analysis of classic pratical network coding. Batch-based random linear network coding (BB_RLNC) can improve network throughput in the wireless network, but it has the extremely challenge to decide the optimal time to move to the transmission of the next batch. To improve network throughput advancely, Sliding window-based random linear network coding (SW_RLNC) was proposed. Yet there is not any research about the affect of sliding window size and step size to the network performance such as throughput and decode delay. We observed that because of SW_RLNC always add new original packets it will cause a series of original packets can’t immediately decoded, which cause the packets that receive node need to cache increase and the decode matrix and decoded complex increase.This paper analysis the throughput and packet decode delay of SW_RLNC and BB_RLNC in lossy and lossy-free network environment. And the results show that when the parameters was selected properly, the throughput and packet decode delay can get better performance than BB_RLNC. The relationship of network performance and SW_RLNC parameters will given when the thourghput and decode delay was optimal and the simulation results validated the theoretic analysis.To validate the performance of SW_RLNC when considering the affect of different layer, we implement BB_RLNC and SW_RLNC in Nsclick simulator and get the performance of them. The results show that SW_RLNC may not outperformance than BB_RLNC as thoritical results. To validate the affect of ACK, we compared two kinds of ACK feedback manner and the simulation results shows that SW RLNC exist serious ACK and data packet collicion and the transmission of the ACK packet will reduce SW_RLNC throughput. So when applicating SW_RLNC to real network application, we can’t simply consider the affect of routing layer, but also the affect of other layer need to conconside. The research result has a good reference value for pracitically applicating network coding.