Research Of Grid Resource Scheduling Management On Quality Of Service

Quality of service (QoS) is an important and hot spot in the study of grid resource scheduling. It involves two aspects: one is assessment models of QoS in grid and the other concerns with how to apply these models in resource or service scheduling. As for the current researches, the study of the two aspects is still in the starting stage. The study of QoS model mainly resorts to the model of quality service put forward in Internet. However, because they are not at the same level and their quality parameters differ from each other, the latter can not be transplanted completely. The studies at home and abroad have put forward some QoS assessment models fit for grid, but they can not be exchangeable as the model of grid resource management itself differs from others. The research findings on algorithms of QoS parameters applied in the resource scheduling are very rare, and we only find Gridsim's algorithms of minimum time and minimum cost in the literature and reports. Therefore, more researches are needed to be carried out all around. This thesis studies QoS assessment models in grid resource management and makes improvements in Deadline and Budget Constrained algorithms (DBC algorithms). The main researches are the following:1. With the help of the hierarchical model of QoS, this thesis puts forward hierarchical mapping methods of QoS: one is the parameter mapping of QoS in one virtual organization; and the other is the parameter mapping of QoS among different virtual organizations.2. The author uses the parameter mapping of QoS to improve DBC algorithms. The author proposes a method which combines k-means and performance optimum algorithm. The central idea is to make loading and distribution more flexible through changing modes of distribution of tasks and to improve the method of optimizing performance through collecting resources of QoS, dividing their grades and balancing the loading.3. Through experiments, the thesis proves that the improved algorithm can balance loading of resources effectively while it does not slow down the fulfillment of tasks.