The Research Of LVS Weight Scheduling Algorithm Based On Adaptive Genetic Algorithm

A cluster system is a group of computers that are made up of multiple nodes that provide high-speed or private networking services and network resources to clients.Each service node is accessed transparently by clients and are independent of each other.The main purpose of a cluster is to enhance performance,control costs,enhance scalability,and improve reliability.According to different users,clusters can be set into separate categories such as a scientific cluster,a load balancing cluster,and a high availability cluster.However,obtaining good performance by a cluster is limited;therefore,to acquire optimal performance,the Linux Virtual Server(LVS)load balancing scheduling needs to be improved.There exist shortcomings within the LVS algorithms that,firstly,set their weight statically.Its weight can only be manually configured by the administrator and cannot be modified once the server is active.In addition,the value of the weight is limited allowing only one server to correspond with one weight.This weight can prohibit the bearing of the actual load capacity of the Real Server(RS).Furthermore,the general improved dynamic scheduling algorithms need to be achieved through real-time communication between the load balancer and the server,increasing their and the network's pressure.In view of the above problems,an LVS weight scheduling algorithm based on adaptive genetic algorithm is proposed.Presenting a new,adaptive genetic algorithm based on the characteristics of the load balancer load scheduling algorithm,real numbers are used as the encoding.This is a combination of a roulette-wheel selection method and an individual preservation method that calculates the fitness value according to the fitness function.This,combined with the arbitrary and trending characteristics of the cloud model,crossover probability and mutation probability can be acquired.This crossover operator of non-uniform crossover and variable step size mutation operator is one way to achieve the desired outcome.According to the operational principle of LVS,the cluster's structure is improved with an increase of the server information acquisition module,which is used to obtain the server integrated load value.Additionally,an increase of the load balancer information processing module,that adapts to the server's comprehensive load value,calculates the fitness value to then calculate the adaptive crossover probability and the adaptive mutation probability.The optimal solution is changing the weight of the server,in coherence to the genetic operation.Finally,in order to verify the new algorithm proposed,the VS/DR deployment method to build a test platform in Linux environment needs to be used to conduct an experimental test which has already been done.The results showed that,compared with the traditional LVS scheduling algorithms,the throughput of the new algorithm has been significantly improved while keeping the response time relatively downscaled.Compared to other comparable dynamic scheduling algorithms,the shortcomings of real-time communications are overcome and the network pressure is relieved.The LVS weighted scheduling algorithm,based on the adaptive genetic algorithm already proposed,remedies the shortcomings and deficiency of the original scheduling algorithms and the third-party dynamic scheduling algorithms.Improvement of the system throughput and the evident reduction of the system response time makes this LVS cluster system's performance a favorable one.