Advance reservation can ensure the future availability of the Grid's heterogeneous resources and help a scheduler to produce better schedules. We propose a new resource mapping technique for jobs of a Grid workflow in an advance reservation environment. Using the Dynamic Critical Path for Grid (DCP-G) scheduling algorithm's job selection method, our proposed technique considers the conditional mapping of parent and child jobs to the same resource, trying to minimise the communication duration between jobs and thus optimising workflow completion time. For a number of different workflows, the schedules computed by our technique are compared with those computed by DCP-G, a workflow scheduling algorithm that is known to outperform many other popular scheduling heuristics. The results show that our proposed resource mapping technique yields improved schedules along with the significant improvement in the CPU time as the size of the workflow increases.

Abstract-Automated advance reservation has the potential to ensure a good scheduling solution in computational Grids. To improve global throughput of Grid system and enhance resource utilization, workload has to be distributed among the resources of the Grid evenly. This paper discusses the problem of load distribution and resource utilization in heterogeneous Grids in advance reservation environment. We have proposed an extension of Partner Based Dynamic Critical Path for Grids algorithm named Balanced Partner Based Dynamic Critical Path for Grids (B-PDCPG) that incorporates a hybrid and threshold based mechanism to achieve load balancing to an allowed value of variation in workload among the resources in Partner Based Dynamic Critical Path for Grids algorithm. The proposed load balancing technique uses Utilization Profiles to store the reservation details and check the loads from these profiles on each of the resources and links. The load is distributed among resources based on the processing element capacity and number of processing units on resources. The simulation results, using Gridsim simulation engine, show that the proposed technique has balanced the workload very effectively and has provided better utilization of resources while decreasing the workflow makespan.

No abstract