A Deadline and Budget Constrained Cost-Time Optimization Algorithm for Scheduling Dependent Tasks in Grid Computing

Computational grid has a promising future in large-scale computing, because it enables the sharing of widely distributed computing resources. Good managements with excellent scheduling algorithms are in great demand to take full advantage of it. Many scheduling algorithms in grid computing are for independent tasks. However, communications are very common in scientific computing programs. In this paper, we will propose an easy-implemented algorithm to schedule the tasks with some communications. Our algorithm is suitable for a large proportion of scientific computing programs, and is based on Binary Integer Programming. It is able to meet the users’ quality of service (QoS) requirements, and to minimize the combination of costs and time consumed by the users’ programs. We will give an example of scheduling a typical scientific computing task to show the power of our algorithm. In our experiment, the grid resource consists of an SGI Onyx 3900 supercomputer, four SGI Octane workstations, four Intel P4-2.0GHz PCs and four Intel P4-1.8GHz PCs.