Load balancing task scheduling algorithm in Hadoop platform

As the cloud platform, it must be the platform with high performance, high reliability, easy expanding, but if only by constantly updating the server to realize the cloud platform, it will undoubtedly exist the following problems: 1, the update of server causes the waste of hardware; 2, the cost of many high performance server as backup is too high; 3, the general design often has network bottleneck, making the performance not play out. The design of this paper using cluster technology of load balanced, to realize the cloud platform reconciling credibility, performance and cost. The network structure that combining the ring network and star network, diversifies the combining form of computation, it also can improve the performance, and the multiple use of load balancing and backup makes performance and safety be reflected. The load balancing technology with consideration of the bottleneck problem of network, makes customer's visit will not only focus on a certain endpoints, which makes the communication process will not meet the bottleneck. Finally, through the analysis of the reliability and performance and the test of the practical application, we found that the performance of the design was better in the case of higher load, and the platform could maintain this state for long term.