Title :
A proposal for a heterogeneous cluster ScaLAPACK (dense linear solvers)
Author :
Beaumont, Olivier ; Boudet, Vincent ; Petitet, Antoine ; Rastello, Fabrice ; Robert, Yves
Author_Institution :
LIP, UMR CNRS-ENS, Ecole Normale Superieure de Lyon, France
fDate :
10/1/2001 12:00:00 AM
Abstract :
The authors study the implementation of dense linear algebra kernels, such as matrix multiplication or linear system solvers, on heterogeneous networks of workstations. The uniform block-cyclic data distribution scheme commonly used for homogeneous collections of processors limits the performance of these linear algebra kernels on heterogeneous grids to the speed of the slowest processor. We present and study more sophisticated data allocation strategies that balance the load on heterogeneous platforms with respect to the performance of the processors. When targeting unidimensional grids, the load-balancing problem can be solved rather easily. When targeting two-dimensional grids, which are the key to scalability and efficiency for numerical kernels, the problem turns out to be surprisingly difficult. We formally state the 2D load-balancing problem and prove its NP-completeness. Next, we introduce a data allocation heuristic, which turns out to be very satisfactory: Its practical usefulness is demonstrated by MPI experiments conducted with a heterogeneous network of workstations
Keywords :
computational complexity; linear algebra; message passing; operating system kernels; optimisation; resource allocation; workstation clusters; 2D load-balancing problem; MPI experiments; NP-completeness; cluster computing; data allocation; data allocation heuristic; data allocation strategies; dense linear algebra kernels; dense linear solvers; heterogeneous cluster ScaLAPACK; heterogeneous grids; heterogeneous networks of workstations; heterogeneous platforms; homogeneous processors; linear system solvers; load-balancing problem; matrix multiplication; numerical kernels; numerical libraries; two-dimensional grids; unidimensional grids; uniform block-cyclic data distribution scheme; Computer networks; Distributed computing; Kernel; Libraries; Linear algebra; Linear systems; Pervasive computing; Proposals; Scalability; Workstations;
Journal_Title :
Computers, IEEE Transactions on