The Design And Implementation Of Scheduling Module With Resource Preference In Grid Computing

The early products of grid computing technology focus on high performance scientific computing, and now the grid has been applied in different fields of grid technology. One of the grid's biggest benefits is on a variety of computing and data resource sharing. The grid as a single system, its essence is the use of grid computing resources and to achieve information sharing. Data sources in grid are various, which have difference in store forms. That information composes attribute information of computing resources in grid system. Convenient and effective for the attribute information in the future will be a major grid middleware product features and key technologies.This article is based on existing grid middleware products, modifies and expands its scheduling policies, so that this product can provide more effective, reasonable, and better scheduling decisions, resolve the problem of task relevant attributes for resources in the specific needs can not be met, help them use their systems most effectively, satisfy their customs' requierments and reduce the possession of their computing resources and cost of management.This article follows the existing architecture of grid computing products, using its original task scheduling strategy for the development of a resource preference scheduling strategy, the grid computing platform to achieve mission based on the specific resource requirements for reasonable the corresponding computing nodes allocated, including specific resources for the compute nodes to collect the published subscriber attribute model to develop plug-in the form of attribute information for real-time calculation data, also contains a reasonable and easy to use user interface and programming configuration interface module. Publisher and subscriber model can eliminate the definition and maintenance of a large number of properties set for a particular resource management, thus simplifying system management, reducing the coupling between system modules, and improving system performance and stability.Finally, with co-ordination of the task scheduling algorithm and properties of various resources, complete the scheduling model with resource preference using the style same as the original product. The scheduling strategy with resource preference can dynamically configure to achieve a better allocation of compute nodes by resource properties.