Research On Congestion Control Forwarding Algorithms For Internet Concurrent Multipath Mechanism

Due to the unbalance distributions of network resource and traffics, the occurrence of congestion is an intrinsic characteristic of the Internet. With the rapid growth of network scale and the emerging of the variety of applications, network congestion problems is growing, which becomes the main factor to affect the network performance. TCP congestion control mechanism based on source nodes is effective in the prevention of congestion collapse, while it also faces many new crises, such as it doesn’t fit for the multimedia application with UDP protocol. Because a router is able to catch the packet information, detect the congestion degree and analyze the effect of the traffic on congestion, while it can balance the network resource and traffic by multipath routing mechanism, so, congestion algorithms for Internet concurrent multipath have been the hotspots in the network research field.In this thesis, we analyze the cause of network congestion and it’s characteristic, and introduce the summary of research status. Based on the analysis of the traffic over a router, we find the main factor to affect the network performance and establish the model of the traffic prediction, then formulate algorithm to solve or avoid congestion.Internet has no warning method to the link failure and burst traffic which will decrease the reliability and validity of the network and affect the quality of service, so we have suggested a packet loss rate prediction algorithm over multipath network. Basis of taking the packet loss rate of the given flow over each and every path as observing value through relevant measurement, the exponential smoothing prediction is introduced. The prediction algorithm has good precision in calculation and can be applied to smooth flow. To increase the precision in calculation, the prediction loss rate algorithm based on support vector machine is proposed, which calculates the linear regression function by regression estimate of the training sample and predicts the loss rate with the function. Compared with the exponential smoothing prediction algorithm, it has better precision in calculation.Because load balance over multipath network is an effective method to solve and avoid network congestion, we have suggested a load balance algorithm based on exponential smoothing prediction. In this method, the application which is sensitive to packet loss is considered. Basis of exponential smoothing prediction introduced to predict the packet loss rate, the load balance algorithm leads the flow to the paths at low packet loss rats and increases the performance of the network.The aim of load balance is to satisfy the quality of service of the given application, such as minimizing the packet loss rate, while it must be meet the restrict of the network resource, so we have suggested a load balance algorithm based on PSO which solves the network problem by optimization problem. The first order exponential smoothing prediction is introduced into the scheme to distribute the packets over multipath in accordance to the PSO algorithm to minimize the network loss rate. The algorithm increases the quality of service and improves the network performance.In order to study the load balance algorithm under the burst background traffic, we have suggested a load balance algorithm based on ARIMA prediction and PSO. Packet loss rate which is considered as knownquantity in the function of the sum of link utility and packet loss rate is predicted by ARIMA algorithm. The load balance rate is calculated by PSO algorithm. The load balance algorithm based on ARIMA prediction and PSO not only improves the quality of service but also avoids the surge of traffic which is the cause of the algorithm which is only considering the quality of service of the application but not considering the affecr of the link utility.To tackle the problem of concurrent flows running, we have suggested a load balance algorithm serving isomorphic flows. Based on analyzing the requirement of each flow, we have transformed the network problem to the optimization problem and used PSO algorithm to solve. The algorithm improves the performance of the network, such as packet loss rate, delay and throughput.Internet is the presence of both elastic and inelastic flow, thus it is imperative that an efficient strategies to jointly manage these two types of flow should be developed. So we have suggested a load balance algorithm in multipath network serving heterogeneous flows. We have formulated a new load balance problem that incorporates the performance requirements of inelastic flow with reliability and elastic flow with network utility. The first order exponential smoothing prediction is introduced into the scheme to distribute the packets over multipath in accordance to HPSO (hybrid particle swarm optimization) algorithm to minimize the objective function. Our approach algorithm not only achieves the maximization utility gained by elastic flow but also satisfies the reliability demand of inelastic flow.