Research And Implementation Of LVS Cluster Load Balancing Scheduling Algorithm Based On PSO-GA

People's life and work are closely connected with the internet,the proportion of network activities in daily life is increasing,so a lot of data is generated.So much data puts huge pressure on the server cluster,this raises the issue of resource allocation.The more reasonable the resource allocation,the more balanced the load of the server,the higher the cluster processing efficiency,the lower the latency,and the higher the user satisfaction.Therefore,it's of great practical significance to study the problem of load balancing.The existing load balancing algorithms take relatively simple factors into account and do not comprehensively consider the server load and the resource consumption of the requests.In view of these situations,this thesis proposes a load balancing scheduling algorithm for LVS(Linux Virtual Server)Cluster Based on PSO-GA(Particle Swarm Optimization-Genetic Algorithm),which quantifies the impact of different scheduling schemes on the current cluster load by constructing a resource balancing model and fitness function,PSO-GA algorithm is used to solve the fitness function to obtain the optimal weights.The load balancer schedules requests according to these weights,so as to realize the Linux virtual server cluster load balancing.The main works of this thesis as follow:1.Based on the resource consumption of the request and the real-time load of the server nodes,this thesis establishes a resource balance model and designs a fitness function.2.On the basis of particle swarm optimization,the mutation idea of genetic algorithm is introduced to form PSO-GA algorithm.This thesis adjusts the inertia weight of PSO-GA algorithm,designs elimination mechanism to eliminate poor individuals,and controls the variation rate.The performance of PSO-GA algorithm is tested by Matlab.3.This thesis implements the whole load balancing algorithm framework on server cluster.This thesis compiles load information collection module and communication module to realize load information collection and transmission of nodes,and compiles PSO-GA calculation module and weight transfer module to realize the calculation and transfer of weights,and compiles scheduling module to realize the application of weights.4.This thesis tests the new load balancing algorithm on the server cluster.The test results show that in the case of large amount of concurrency,the load balancing algorithm proposed in this thesis performs better than the other three load balancing algorithms in terms of throughput,response delay and request error rate,and is more balanced in terms of resource utilization.