Grid computing is an emerging computing paradigm and is distinguished from distributed computing by its efficient and optimal utilization of heterogeneous, loosely coupled resources tied to work load management. However, complexity incurred in efficient management of heterogeneous, geographically distributed and dynamically available resources has become one of the most challenging issues in grid computing. A lot of parameters have to be taken into consideration to efficiently utilize the grid resources. Many heuristics has been proposed in the literature to address this complex problem. Present research aims to solve load balancing decisions using Artificial Neural Networks (ANN). Since ANN are best at identifying patterns or trends in data, their ability to learn by examples makes them very flexible and powerful. In this research we have developed and evaluated a completely new scheduling-cum-load balancing module for a scaleable grid. Experimental results suggest that once trained, ANN outperforms other heuristic approaches for large tasks. However for small tasks, ANN suffers from extensive overheads.
展开▼
