Research On Adaptive Congestion Control Algorithm Based On Link Awareness

With the development of new network structures and the arrival of the 5G communication era,the traditional single network structure is transformed into a complex heterogeneous network structure combining multiple networks.The expansion of network scale and the surge of data traffic make the network congestion more and more serious.In the current complex heterogeneous network with high bit error rate and other characteristics,the loss of data packets is not only caused by network congestion,but also caused by random bit errors caused by link signal interference.The traditional TCP congestion control mechanism lacks self-adaptive perception of the link status in the network,and cannot effectively judge the type of packet loss,resulting in unnecessary congestion avoidance behavior.Similarly,the traditional router-based network layer congestion control algorithm is sensitive to parameters and lacks adaptability.In complex and changeable network scenarios,once the transmitted data traffic fluctuates greatly,the network performance deteriorates sharply,resulting in poor stability and other problems.Therefore,it is necessary to systematically design the global control of network congestion by considering various factors in the network.The main research contents of this thesis are as follows:Firstly,in order to deeply study the principle of network congestion mechanism,this thesis studies the classical algorithms of TCP congestion control based on the source end and network layer congestion control based on the router end respectively,to analyze the factors that cause network congestion,and the relationship and defects between the two congestion control mechanisms.Secondly,this thesis proposes an adaptive congestion control algorithm based on link sensing,namely TCP NRBE algorithm.The algorithm senses the link state by predicting the congestion level of the packets,then estimates the available bandwidth to judge the type of packet loss,and realizes the adaptive adjustment of the congestion window by real-time link state.The simulation results show that the algorithm is superior to the existing TCP congestion control algorithms based on packet loss and delay in terms of congestion window growth,throughput,fairness,convergence and friendliness.High communication throughput and competitive fairness are achieved in the network,and can be widely used in complex heterogeneous network applications to achieve high-performance data transmission without modifying any hardware in the network.Finally,based on the proposed link-aware-based adaptive congestion control algorithm,the TCP NRBE-RED continuous-time model is further established to construct a congestion control system.The two parts of the entire network system are based on the source-side and router-side congestion control.Modeling separately and cascading as a whole make the two congestion control mechanisms cooperate and interact with each other,build a complete network closed-loop control system,and complete the management of network congestion.The classic automatic control theory is used to analyze and optimize the queue management parameters of the RED algorithm to realize the self-adaptation adjustment based on link awareness and solve the problem of sensitive parameter settings.Through theoretical modeling analysis and NS3 simulation experiments,the effectiveness of the research strategy is verified,and the smooth and efficient operation of the network is ensured.