Research On End-To-End Congestion Control Mechanism Of Low-Orbit Satellite Network

The forward development of society is inseparable from continuous research in the field of communication.After realizing the need for low latency and high bandwidth,people began to pursue a new generation of networks that integrate heaven and earth and cover the whole world.So let the low-orbit satellite network,after slow progress in previous decades,reenter the public eye and be vigorously promoted by governments.However,due to the influence of the transmission channel on shading,rain decay,multipath,etc.,the low-orbit satellite network has a high bit error rate.At the same time,the physical distance of communication,the switching and interruption of links lead to long transmission delays.These characteristics directly damage the reliable transmission of data and reduce the throughput performance of the existing congestion control mechanism,so this paper designs a new congestion control algorithm according to the characteristics of low-orbit satellite networks:1.The BBRN(BBR-NEW)algorithm based on stage optimization is proposed.The BBR algorithm is a congestion control algorithm developed in recent years.It probes the resources available to the network by asynchronously measuring the network bandwidth and the minimum latency.Control the send rate based on the estimated bandwidth,trying not to occupy the bottleneck queue,so keeping latency low while pursuing high throughput.But the BBR has two problems when applied to low-orbit satellite networks.The first problem is slow convergence time.Poor convergence time is due to a combination of insufficient pacing rate and long stage switching times.For insufficient pacing rates,the reason for the analysis is due to the small growth factor in the startup phase.Therefore,the theory derives the conditions that the growth factor meets,optimizes the production of new growth factors,and improves the response speed of the startup stage.For excessively long switching times,the reason for the analysis was due to the fact that BBR required 3 valid confirmations of RTTs.Therefore,by introducing the GM(1,1)grey prediction model and designing the transition propulsion module,the two-stage conversion process was accelerated.The second problem is that bandwidth measurements are inaccurate,and the analysis for this problem is due to a lack of processing for unusual bandwidth values.Therefore,the bandwidth processing module is introduced,which improves the accuracy of the available resource detection.Thus,the BBRN algorithm based on stage optimization is formed.In the NS3 simulation,it was found that the new growth factor of BBRN can shorten the convergence time by 22%,shorten the time of switching the drain stage in the startup stage by about 1 RTT,and the recognition rate of abnormal bandwidth values is more than 80%.Compared with BBR and RBBR,BBRN’s bandwidth utilization increased by 14.6%and 15.1%respectively under the condition of large packet loss rate.2.A BP(BBR-Prediction)algorithm based on predictive perception is proposed.Because of the long latency and large packet loss characteristics of low-orbit satellite networks.Therefore,there are problems such as weak network perception and inability of congestion control strategies to adapt.For weak perception capabilities,it is first necessary to select reasonable network resource characterization parameters.This article abstracts network resources into two types of caches,measured in terms of bottleneck bandwidth and RTT.In order to compensate for the hysteresis of bottleneck bandwidth measurement,EKF-based bandwidth prediction is proposed with the help of EKF technology.In order to eliminate the noise during the RTT measurement and obtain the effective queuing delay,the STL-based RTT decomposition is proposed by using STL technology.Aiming at the problem of poor adaptation of the existing control strategy,the new congestion control function is redesigned according to the processed bandwidth and RTT,so as to form a congestion control strategy based on predictive perception.Finally,different network properties are set in the simulation environment.Simulation results show that the average latency of the BP algorithm is reduced by 52.4%compared with the packetloss algorithm Cubic,the effective throughput increases by 5.8%compared with the delay algorithm Vegas,and the packet loss rate decreases by 29.7%compared with the BBR algorithm.Compared to other algorithms,the performance has also improved.