Task scheduling using performance-driven

Computational grids aim to aggregate the power of heterogeneous, geographically distributed, multiple-domain spanning computational resources to provide high performance or high-throughput computing. To achieve the promising potentials of computational grids, an effective and efficient scheduling system is fundamentally important. Grid is also a dynamic resource-sharing computing environment in which there are many applications running. At time of scheduling decision making, accurate resource predication on run-time resource measurement parameters is critical to maximize application performance. In this paper, a method for estimating task running time and two scheduling algorithms are presented to facilitate effectively dynamic scheduling in a heterogeneous grid environment and improve application performance.