Fairness Optimization Research Of TCP-BBR Congestion Control Algorithm

With the diversified development of the Internet,the number of various applications,especially audio and video applications,have increased substantially.The traditional TCP congestion control algorithm cannot meet the high network transmission requirements of current Internet applications.In this context,Google proposed a hybrid congestion control algorithm based on Bottleneck Bandwidth and Round-trip Propagation time(BBR).BBR congestion control algorithm measures the network bottleneck bandwidth and minimum delay in real time,and adjusts the sending rate by calculating the bandwidth delay product to achieve high throughput and low delay.Therefore,BBR algorithm is regarded as a new era of congestion control algorithms.However,relevant studies show that BBR still has some problems,such as unfairness,high retransmission rate,deep buffer performance degradation and so on.Therefore,we need to do more researches on its performance and constantly improve the BBR congestion control algorithm.This dissertation studies the working mechanism of BBR congestion control algorithm,analyzes and optimizes its most prominent fairness problem,in order to improve the congestion control performance of the algorithm.For the Round-Trip Time(RTT)fairness,we discuss the fairness optimization scheme from two aspects:Congestion Window(CWND)and sending rate,respectively.A fluid model is used to describe the data transmission process in BBR congestion control.On the basis of the fluid model,the parameter optimization strategy of cwnd gain and pacing gain is proposed to improve the fairness of the algorithm and other network transmission performance.Moreover,we evaluate and and optimize the latest version of BBR(BBRv2).This dissertation provides effective optimization methods and suggestions for the fairness problem of BBR congestion control algorithm in the application,so as to provide the scientific and effective reference for the evolution of BBR algorithm and the design of the next generation TCP congestion control algorithm.The main research work can be summarized as follows:Firstly,the working mechanism of BBR congestion control algorithm is studied,the fairness of BBR algorithm is analyzed and discussed.Furthermore,the root causes of its fairness are analyzed and studied through literature research and simulation experiments.At the same time,it is found that although the inter-protocol fairness of BBRv2 algorithm is improved,the RTT fairness and intra-protocol fairness problems are still obvious.Secondly,the influence of CWND on RTT fairness in BBR algorithm is analyzed,and an optimization algorithm based on adaptive CWND is proposed.According to the delivery rate and RTT law,the cwnd_gain is adjusted by self-defining adjustment factors a and β,which alleviates the limitation of CWND on short RTT flows and reduces the CWND advantage of long RTT flows.Simulation results and real environment tests show that the optimization algorithm can effectively improve the RTT fairness of BBR algorithm,and improve the performance of channel utilization and retransmission rate.Thirdly,the detection mechanism of sending rate of BBR algorithm is studied.Due to the asynchronous exploration of bottleneck bandwidth,the sending rate does not match the bottleneck bandwidth,resulting in RTT unfairness and high delay,an optimization scheme based on pacing gain model is proposed.Instead of using a fixed pacing gain coefficient,the sending rate is adjusted by elastic interleaving the up and down pacing gain coefficients so that each BBR flow can compete fairly for bandwidth resources.According to the variation law and complexity of pacing gain,we analyze and select three different pacing gain regulating functions.The evaluation results of simulation and real network scenarios show that the RTT fairness and other performance of BBR algorithm are improved effectively.Fourthly,this dissertation studies and analyzes the latest BBRv2 algorithm.According to the mechanism of packet loss rate and Explicit Congestion Notification(ECN)marking in bandwidth detection,a flow-aware ECN is proposed.by quantifying queue information and congestion degree,the method adjusts the threshold judgment of packet loss according to the congestion degree,to avoid the blind CWND restrictions.At the same time,ECN marking is done selectively according to the queue information,slowing down the speed of the high-speed flow,accelerating the completion of the lowspeed flow,so as to achieve the balance of the sending rate.The experimental results of simulation and real environment show that this optimization algorithm can effectively alleviate the fairness problem in BBRv2 algorithm and improve congestion control performance.