Research On MPTCP Buffer Based On Regression Prediction Statistics Model

With the continuous evolution of MPTCP(Multipath-TCP Multipath Transmission Control Protocol),people have noticed that the robustness,throughput,and transmission performance of the MPTCP protocol are continuously enhanced,but the calculation of the buffer consumption of the multipath transmission is still in progress.A simple extension of the single-path buffer consumption calculation formula is followed,however,only the lower limit value of the buffer consumption is calculated through this simple extension formula,which is rather large and wastes resources.Obviously,this method of simply expanding a single path can not really solve the problem of buffer consumption in actual multipath transmission.Therefore,this paper introduces a mathematical method to study the usage of MPTCP buffer area.The specific method is to use mathematical statistics analysis to find The relationship between buffer,throughput,and congestion control algorithms.At present,the MPTCP protocol has accumulated a lot of theoretical knowledge,but there are only some qualitative discussions and descriptions on how to set the appropriate MPTCP buffer area by means of mathematical statistics.However,there are relatively few specific quantitative analysis and modeling studies.Because the MPTP protocol buffer consumption of multi-path transmission is very different from that of the traditional TCP,the traditional TCP single path transmission cannot be used to treat and solve the problem of multi-path transmission.Therefore,for the first time,the idea of mathematical statistics is The problem of multi-path transmission buffer consumption is solved by mathematical thinking to solve the problem of dynamic calculation of its buffer consumption.As we all know,the mathematical statistics method has a wide range of applications,and it has been successfully applied in many industrial and other industries.This paper has conducted an exploratory study of the problem through this method,carried out a large number of test scene data analysis,and built Modulus,the results show:1.The throughput of the sender and receiver is a quadratic function of the buffer value of all paths;2.different congestion control algorithms affect the throughput;3.Since the choice of the initial path of the multi-path protocol has no obvious effect on the throughput,the initial path factor is excluded when performing statistical analysis modeling.These conclusions are all derived from the consideration of network parameters of different scenarios,such as buffer values,congestion control algorithms,and network provider ISPs.The specific research process involves test scenario design,test script design,data distribution and recovery script design,selection of potential influence factors for throughput,establishment of a regression analysis relationship model,and a large number of test experiments.Finally,mathematical formulas are used to model the equations.In addition,the prediction model of multi-path transmission buffer consumption obtained in this paper shows that the size of the buffer for handling network traffic can be dynamically determined by the throughput size,while the traditional TCP protocol uses a static method to obtain a fixed buffer size;Through this prediction model,it is calculated that when the network traffic is low,the buffer consumption is much smaller than the buffer value obtained by the traditional multi-path transmission buffer consumption calculation method.