Research On The Design And Algorithms Of A High Availability Web Cluster Model Based On Prediction

In the research field of Web clusters, how to perform admission control on the requests, distribute the requests, and schedule the request tasks in back-end servers, have become hot research topics. The traditional solutions always focused on certain mechanism. The research aim of this paper is to propose a new Web cluster model that can provide admission control, request distribute, and task schedule mechanism.In the thesis, some traditional Web cluster models was introduced, and several representative admission control, request distribute, and task schedule algorithms are analyzed, then their advantages and disadvantages are concluded. Based on the analysis, a new Web cluster model and associated algorithms are proposed. The architecture of the new model and its mechanism are explained clearly, and the associated three algorithms, admission control, request distribute, and task schedule, are discussed comprehensively. At last, the proposed mechanisms are compared with some similar mechanisms, the result demonstrates that they have high system resource utilization and throughout, and are much better than other mechanisms in session integrity and value realization ratio.The model and algorithms are rely on the predication of system resource utilization and task run time, and focus on the session integrity and service differentiation, hence the Web cluster can have a high throughout, and provide better Web QoS guarantee. The model is very suitable to modern Web application, especially e-commercial, e-government, so it make sense in both theory and practice.