Dynamic matching of independent jobs on a computing grid

Grid Computing has emerged as a new paradigms for distributed computing. It promotes sharing of distributed resources that may be heterogeneous in nature and enables scientists and engineering professionals to solve large scale computing problems. With grid computing, businesses can efficiently utilize computing and data resources and combine them for large capacity workloads. The primary benefit of grid computing is the ability to coordinate and share resources.; To maximize the benefits of grid computing, it is essential to be able to discover the resources available on the grid, and also be able to map the jobs to the resources most effectively to maximize the given objective function. In computational grids, the problem of resource discovery and mapping (matching and scheduling) of jobs to resources is really challenging due to the large number of resources, resource heterogeneity and variation in resource attributes such as CPU load, network bandwidth etc. This research focuses on the problem of dynamic matching of jobs to resources in a computing grid. The state of the art resource discovery and dynamic mapping policies are studied in detail. Most of the mapping policies assume complete knowledge about the characteristics of jobs and resources while making an allocation decision. However, this may not be feasible in reality. This research introduces dynamic matching policies based on job classes, which match jobs to resources on a grid without any detailed knowledge about the characteristics of jobs and resources. Under a variety of different workload parameters they demonstrate a performance comparable to or better than the minimum completion time mapping heuristic that is based on exact a priori knowledge of job and resource characteristics.