Immediate mode scheduling of independent jobs in computational grids

Abstract

With the emerging paradigm of grid computing and the development of grid infrastructures, grid-based applications are becoming a common approach for solving many complex, large-scale problems in science and engineering. In order to benefit from the large computing power of grid systems, efficient allocation of jobs to resources is necessary. In this work, we consider the allocation problem in immediate mode, in which jobs are allocated as soon as they arrive in the system. We implemented several methods and measured four parameters of the system: makespan, flow- time, resource utilization and matching proximity. The immediate methods are especially interesting when good quality allocations are necessary in very short time. The considered methods have been tested using the most difficult benchmark in the literature for the problem. The computational results allowed us to identify which of considered methods perform better for makespan, flowtime, resource utilization and matching proximity. Also, we were able to evaluate the usefulness of such methods if we knew in advance certain grid characteristics such as degree of consistency of computing, heterogeneity of jobs and resources.