Font Size: a A A

Research On Dynamic Load Balancing Technology Of Multi-WAN Router

Posted on:2022-06-11Degree:MasterType:Thesis
Country:ChinaCandidate:P YangFull Text:PDF
GTID:2518306734986259Subject:Master of Engineering
Abstract/Summary:PDF Full Text Request
With the continuous updating and iteration of science and technology,the rapid increase of network users and the improvement of network service quality requirements,the single WAN port router can not well meet the increasing network requirements of users.The solution of most enterprises is to introduce multi-WAN routers,among which load balancing is the key to the research of multi-WAN routers,which can effectively improve the throughput of routers and reduce the average response time.The topic of this paper comes from the product upgrade of a company,and the main research content is as follows:(1)A real-time dynamic load weight balancing algorithm is designed.Periodically collect delay,jitter,packet loss rate,and bandwidth usage to comprehensively measure link performance.Set load weights to modify thresholds and compare them with the standard deviation of the average response time of the two time points to determine whether to recalculate the link load weights at this time point.The weight coefficient of the evaluation function was calculated by combining the improved analytic hierarchy process(AHP)and the standard deviation method.The real-time link load weight was calculated by the evaluation function designed by the weighted sum method,and then the user request was allocated according to the link load weight.At the same time,the dynamic determination of acquisition period is realized through experiments.This algorithm can improve the throughput of the router and reduce the average response time.(2)An optimization algorithm of real-time dynamic load weight balancing algorithm based on BP neural network is designed.In order to optimize the real-time dynamic load weight balance algorithm,the adaptive characteristics of the BP neural network are used,and the Adam optimization training function is used to realize generalized learning.Simulation experiments prove that this algorithm can reduce system delay and response time,and improve the load balancing effect.(3)In the Linux environment,a multi-WAN router is constructed and the algorithm is verified by simulation experiments.The average response time and throughput were selected as the performance indexes of the algorithm by JMeter test tool.The simulation results show that the multi-WAN port has a lower average response time than the existing single WAN port.As the number of concurren requests increases,the proposed algorithm has lower average response time and higher throughput compared with Smooth Weighted RoundRobin and Weighted Least Connection algorithms,and the load balancing effect is best when the number of concurrent requests is 1500.
Keywords/Search Tags:multi-WAN router, Dynamic load balancing, Real-time dynamic load weight balancing, BP neural network
PDF Full Text Request
Related items