Co-scheduling Algorithm Based On Collaborative Reserved

Grid computing realized the sharing of resources of variety locations of homogeneous or heterogeneous. The core of Grid Computing is how to manage the shared resources, while the allocation of resource is an important resource management issue, particularly the assignment of more than one resources to a task. This is co-allocation. By the decomposition of a task into multiple subtasks and then assign these subtasks to multiple resources to execute simultaneously, the co-allocation technology meets users' specific performance requirements. And Co-reservation ensures several resources allocate to one job at the same time when Co-allocating.In this paper, two models,task model with deadline limit and rectangular available resource model, are proposed. And schedule algorithms for Static-immediate deadline task and Static-general deadline task are studied. Also, this paper put forward seven co-reservation algorithms for Static-general task, which are First_fit,PE_bestfit,PE_worstfit,Duration_bestfit,Duration_worstfit,PEDuration_ bestfit and PEDuration_worstfit based on the different total of free PEs , duration of free time and area of free rectangular available resource for all machines which is selected by task on different time.Experiments show the impacts of reservation ratio and reservation time on the average waiting time, the average expansion factor, the weighted average response time and the success rate of reservation when co-reservation Static-immediate deadline tasks. When co-reservation Static-general deadline tasks, the experiments test different resource reservation algorithms, deadline factors, reserve ratios and reserve time factors and show the different impact on the average waiting time, the average expansion factor, the weighted average response time and the success rate of reservation. The result shows the reservation time on the average waiting time, the average expansion factor and the weighted average response time for co-reservation is minimum when co-reservation algorithms is First_fit, and the success rate of reservation for co-reservation is maximal when co-reservation algorithms is PE_worstfit.