Study of load balancing strategies for finite element computations on heterogeneous clusters

  • Authors:

  • Kalyani Munasinghe;Richard Wait

  • Affiliations:

  • Dept. of Computer Science, University of Ruhuna, Sri Lanka;Dept. of Scientific Computing, Uppsala University, Sweden

  • Venue:
  • PARA'04 Proceedings of the 7th international conference on Applied Parallel Computing: state of the Art in Scientific Computing
  • Year:
  • 2004

Quantified Score

Hi-index 0.00

Visualization

Abstract

We study strategies for redistributing the load in an adaptive finite element computation performed on a cluster of workstations. The cluster is assumed to be a heterogeneous, multi-user computing environment. The performance of a particular processor depends on both static factors, such as the processor hardware and dynamic factors, such as the system load and the work of other users. On a network, it is assumed that all processors are connected, but the topology of the finite element sub-domains can be interpreted as a processor topology and hence for each processor, it is possible to define set of neighbours. In finite element analysis, the quantity of computation on a processor is proportional to the size of the sub-domain plus some contribution from the neighbours. We consider schemes that modify the sub-domains by, in general, moving data to adjacent processors. The numerical experiments show the efficiency of the approach.