The Design And Implementation Of Load Balancing Algorithm Based On Cluster

With the development of science, there have been a large number of applications demanding high computing power, while the high-cost of high-performance computers prompted the birth and development of the cluster. Clustering is based on network technology to connect a workstation or a combination of PC, it is a unified integration of resources in the work, and all the nodes use a single interface. It can complete high-performance computers work, but is relatively inexpensive.But it should be some cluster management systems to manage these machines, as machines only connected by the network cannot form clusters. Cluster Management System based on the needs of users, unified manage and schedule the software and hardware resources in the clusters to ensure that user’s job takes a fair and equitable sharing of cluster resources, as well as to increase system resource utilization and throughput.Cluster load balancing is the key to improving the overall performance of the cluster, and its purpose is to reasonably assign the task to the back real service node based on the performance of the node, so that each node achieves a balanced sharing of tasks to improve the performance and service quality.The performance of load balancing algorithm for the cluster design will directly affect the performance of the cluster. Dynamic feedback based on the server real-time load and responding situation, is constantly adjusting the proportion of requests among servers to avoid the situation that some of the server overloaded are still receiving a large number of requests, with the enhancement of processing power of the entire cluster.This article describes the main concepts and features of the cluster. It mainly focuses on the principles and characteristics of load balancing algorithm, describes commonly used strategies and algorithms in the implementation of load-balancing in detail, focuses on the advantages and disadvantages of some commonly used load balancing algorithm, makes further study on the dynamic feedback algorithm and makes some improvements especially on the selection of load parameters and dynamically adjusting of value and finally realize the cluster load-balancing system and make some validity of this method.