Research On Load Balancing Strategy Of Streaming Media Server Cluster

With the rapid development of Internet and computer technology,the 4G network has been popular,people's demand for streaming media service is getting higher and higher.A single streaming media server can't meet the needs of users,and then use streaming media server cluster to solve this problem.Load balancing,as a core technology of clustering,can improve the system performance of the cluster,and its effect usually depends on the design of the load balancing algorithm.Although the traditional dynamic feedback algorithm can obtain the load condition of the cluster node on a regular basis,when the load of a cluster node in the cycle changes greatly,it can't feedback its real load condition in time,and can't better to process a large number of concurrent requests for bursts.In this situation,it is possible to cause the load dump of the cluster.Firstly,this paper introduces the development of load balancing technology and the related protocols of streaming media.Secondly,deeply analyzed and studied the existing static and dynamic load balancing algorithms,and the applicable scenarios of various scheduling algorithms are clarified.Then,the specific improvement ideas are raised for the problems existing in the traditional dynamic feedback algorithm:1.according to the changes of tasks number per second of the cluster node to modify the load feedback cycle dynamically,to enhance the cluster node load feedback timeliness.2.according to the load conditions of the cluster node,they are divided into low load,normal load,high load three categories,according to the total load of the categories to allocate the tasks and using the Least-Connection algorithm allocation task in the class,to solve the problem of load inclination when a large number of concurrent tasks requested.3.using the relay function of streaming media server to achieve load migration of the overload node,and enhance the cluster load balancing effect in the further.Finally,this paper implements the optimization of the load balancing scheduling strategy,and builds a streaming media server cluster to validate the optimal scheduling strategy and compares it with the traditional dynamic load balancing algorithm and the Least-Connection algorithm.The experimental results show that the optimized scheduling strategy can feedback the load condition of the cluster node more timely,and can effectively solve the problem of cluster load inclination caused by a large number of concurrent requests,and improve the load balancing effect of the cluster and the service quality of the cluster.