Fault tolerant scheduling and load balancing for computational grids

Abstract

In the present scenario the computing resources required for solving complex real time applications may not be available at a centralized location due to many practical problems Hence to solve such problems distributed clusters are used which provide unlimited computing resources Heterogeneous clusters consisting of diverse computers interconnected by high speed networks have been adopted as an efficient high erformance newlinecomputing platform for compute intensive applications The limitations of the conventional distributed systems can be handled to fully exploit the underutilized newlinecomputing resources in every area around the world for the distributed jobs Although fault tolerance job scheduling and load balancing are established research areas in distributed computing the grid environments still present new challenges which include large scale computing resources uneven job arrival patterns autonomy of organizations hosting the resources and QoS requirements Grid computing utilizes the distributed heterogeneous clusters to support large scale complicated computing jobs In large scale newlinegrid environments the underlying network connecting them grid resources and processing capabilities of the resources are heterogeneous and bandwidth across resources varies from link to link newline newline