A destination capability aware dynamic load balancing algorithm for heterogeneous environment

Abstract

Complexity of both data- intensive problems and computation-intensive problems has been increased in computational world. These problems require sophisticated mathematical & statistical techniques as well as knowledge creation and involve complex computations to be performed on huge amount of data. A number of large computational problems now work in terms of Petabytes of data rather than limited volume comprising of only Mbytes or Gbytes of data in olden days. In the present scenario, efficient utilization of system resources is necessary to solve large computational problems within desirable time constraint. In this paper, we present a dynamic load balancing algorithm for a cluster created by using heterogeneous commodity hardware. The algorithm uses system resources efficiently & effectively depending on the nature of the applications. Using this algorithm we observed an improvement in execution time of parallel applications if tasks are processed on favorable nodes instead of randomly assigned nodes. © 2011 Springer-Verlag.