Research On Resource Allocation Approach Based On Bayes Policy In Grid

Because of the characteristics of grid resource such as distribution, heterogeneity, dynamic, etc, resource allocation under grid environment is very difficult. Resource allocation is responsible for providing best services according to user's requirement. However, user's request has much fuzzy actually, so quality of grid service is hardly to meet user's needs. Therefore, resource allocation has been one of the key research fields of grid systems.This thesis takes resource allocation as a core, introduces some correlative technologies, characteristics, and current research status in grid computing systemly. After analysising and studying some current popular mechanisms, a Bayes-Based approach for grid resource allocation is presented to solve some key problems in this area.To maximize the economic income of grid system, considering applications' QoS requirements, the approach is based on the idea that price can represent resource's QoS comprehensive performance. By using Naive Bayes Classfication Model, the resource of congruent charge range is allocated to the task according to price parameter that the user submits, and two approaches for QoS level partition are provided, then tasks of higher-level QoS are given a priority, resources of higher-level QoS currently are obligated to tasks of higher-level QoS later, with tasks of lower-level QoS currently being ignored.The experiment shows that the proposed approach can not only meet users' QoS requirements but also make higher economic income for grid system.