| Grid computing, for its advantages and excellent performance, has been attracted worldwide attention from both academic and industry in recent years. The key technologies in Grid computing includes: optical network infrastructure, job scheduling, resource discovery and management, etc. Since job scheduling algorithm is closely related to the efficiency and performance of Grid computing, it plays an important role in the research of Grid computing. Therefore, in this paper, efficient Grid job scheduling mechanisms are studied by taking the characteristics of jobs and resources into consideration.In the first section, the conceptions, architectures and key technologies of Grid computing are introduced. Moreover, the GridSim, which is a famous simulator for Grid computing, is described.In the second section, current Grid job scheduling algorithms are discussed, including Min-min, QoS Guided Min-min, Grouped Task Scheduling Algorithm and Ungrouped Task Scheduling Algorithm, etc. Their performances are discussed in detail in this section.In the third section, two novel job scheduling mechanisms are proposed, including Dynamic QoS-based Task Scheduling with Partial Cloning (DQPC) and Dynamic QoS-based Task Scheduling with Packing and Partial Cloning (DQPPC), by taking the QoS requirements and job relationship into account. The simulations results carried out by GridSim show that the proposed algorithms outperform the traditional algorithm, i.e., QoS Guided Min-min, to a great extent. The final section concludes the proposed algorithm, and discusses the disadvantages of the algorithm and our future work. |
