Optimizing job scheduling in the Grid environment

Job scheduling is essential in achieving effective use of resources in the Grid environment. Scheduling jobs is an optimization problem for some target function. Our goal is to schedule a sequence of independent jobs with the purpose to minimize the total cost. We present a cost model and suggest a locally optimal solution. We find that sending jobs in a good order can improve performance of both the owner machines and the Grid. We also introduce the use of virtual loads to predict changes of workloads of the Grid, and show that it positively impacts performance. Preliminary evaluation is performed using a simulator.