Research And Design Of Task Scheduling Strategy Based On Linux Cluster Systerm

The explosive growth of the Internet applications requires that web servers should have more scalability and more availability. It is the purpose of this paper to find out some way to solve those problems.A single server cannot satisfy the increasing demand. The server cluster, which is a set of autonomous servers connected by a high performance network, becomes the effective way to realize the high scalability and high availability of web server system.Linux is one of the operating systems, which has been rapidly developped recently. With more dependence on Linux servers, building up web servers by using clustering technology based on Linux is one orientation of future web servers.The principle of clustering is to connect a set of autonomous servers by a network as a whole, which offers services to the application requests and distributes those requests to the individual servers in the cluster so as to make workload balanced and to shorten the response time. The workload balancing is related to the performance as well as the scalability and availability of the cluster system. It is the key issue of building up a server cluster.The principle and technology of Linux workload balancing were introduced and some improved algorithm was designed based on existing ones in the paper.The conception, principle and architecture of Linux workload balancing were analyzed in detail.Based on the analysis of existing technologies, a workload balanced Linux system could be built up by using a feedback scheduling algorithm as an extension to LVS, which improved existing eight static scheduling algorithms. As it is well known, the static scheduling algorithms, which depend greatly on the tasks' or servers' static attributes, couldn't distribute the workload evenly among a cluster of back-end servers, due to the tasks' unbalancing and the dynamic environment. The dynamic feedback scheduling requires some mechanism for monitoring and evaluating the present workload on each server so that it could select the right server to undertake the incoming task.The framework and implementation of the dynamic feedback scheduling were included in the paper.Finally, the workload balancing tests had been carried out with the help of load testing tools.The test result showed that the dynamic feedback workload balancing scheduling proposed in the paper had better performance in the workload balancing process than those of the existing algorithms. In addition, it was fault tolerant. And therefore the system reliability could also be improved.