Improvement Of The Calculation Method Of Packet Loss Probability Of RED Algorithm

Since the development of the Internet into the 21st century,the forms of media have become various.The rapid development of various streaming media such as instant messaging and high-definition audio and video has led to an increasing demand for networks.Similarly,with streaming media such as 4K video,pressure on resource providers is increasing,network quality has a great impact,transmission time has not become unstable,and packet loss has become more frequent;On the other hand,in the process of network transmission,when a certain part of the link is congested and the barrel effect is caused,the entire network becomes jammed,that is,network congestion.However,the causes of congestion are often extremely complex.A single control may simply shift the bottleneck,or worse,make the performance worse.The topic of network congestion control is still one of the most important topics at present.Although there are more and more researches in this area,the fundamentals are mainly two major aspects.One is to trigger protocol control from the root,and the other is to use plug control from an intermediate node(usually a router).The router can monitor in real time and can detect the control of various flows on congestion.Therefore,the latter has an absolute advantage in solving the congestion problem.This article also starts from this aspect.The RED(Random Early Detection)algorithm was once an effective solution to network congestion.However,due to the over-abundance of network applications and severe resource competition,current algorithms cannot meet the requirements well.This article starts with the basic concepts,and narrates the principle of congestion control and the current research status.First,propose several concepts,and some commonly used queue management algorithms,then introduce the RED algorithm in detail,analyze the advantages and disadvantages of the algorithm,and then propose an improvement approach to the parameter sensitivity in the RED algorithm.This paper proposes a non-linear algorithm that enhances the probability of packet loss in the original algorithm,replacing the original jump probability curve with a smooth probability curve.Although the principle of the RED algorithm is simple and easy to operate,the number of parameters involved is not easy to set.The newly proposed Sig-RED algorithm draws on the characteristics of the Sigmoid function and changes the formula for the probability of packet loss,using the sigmoid function to calculate the probability of packet loss,to avoid sudden loss of packet loss probability happened.NS-2 simulation experiments show that the Sig-RED algorithm can effectively stabilize the queue length,reduce the influence of the burst queue on the calculation of the average queue length,and reduce the sensitivity of the parameters.There is also obvious optimization in the end-to-end delay,and the performance is better than the RED algorithm.