Study On Dynamic Feedback Load Balancing Algorithm Based On Cluster

Along with the continuous development of information technology, the networkhas gradually become an important part of people’s daily life. And the conventionalservers make it the bottleneck of network services because of the increasing number ofusers and extensions of service types. The logical solution is to use the server cluster.The cluster system has higher performance and dependability as well as extensibility. Itcan enhance the process ability with data and expand the bandwidth of networkequipment and throughput so as to improve the flexibility and reliability of services.The load balancing is the key issue of the cluster, and the core is the load balancingalgorithm. So research with load balancing algorithms has extremely importantsignificance with the clusters development and application.The thesis discusses the current status of the cluster and load balancing andanalyses the system framework of the load balancing cluster and common balancingtechnique and algorithms. Contraposing the problems such as traditional algorithmscan’t estimate the system’s state accurately, the thesis presents the load balancingalgorithm with dynamic feedback based on load weights. The algorithm guides theallocation of missions through the load weights and distribution probability calculatedwith the performance and load of nodes, and improves the balancing efficiency to someextent. It also uses load correction preventing tilts to increase the stability of system. Onthis basis, the thesis presents the load balancing algorithm based on services to solve thebalancing issues of cluster running more services simultaneously. Using differentarguments for different services, the algorithm calculates the performance and load ofnodes to improve the accuracy and guide the mission allocation, guaranteeing that thedistribution is equitable and reasonable. It is hope to make full use of all the resourcesand promote the throughput of cluster. Using OPNET for simulation, the resultsindicates that the algorithm can evaluate the system’s state accurately, make the best ofthe resources, assign the mission to the suitable node and promote the wholeperformance of the cluster. And the simulation tests and verifies the availability of thealgorithm.Based on the research above, the tow algorithms can promote the performance ofcluster system to some extent, and enhance its stability and availability. It has a certainguiding significance for after studies. Moreover, the information technology is developing quickly, and load balancing technique can consider to be combined withnew technologies such as cloud computing to explore a broader development prospect.