On the speedup of distributed linear solvers

Zuberek, W.M. and Perera, T.D.P.

Proc. 5-th EUROSIM Congress on Modeling and Simulation, Special Session on Modeling and Simulation of Distributed Systems and Networks; Marne la Vallee, France, 6-10 September 2004, pp.222-223.

Abstract:

Speedup is usually defined as the ratio of the execution time of an application on a single processor to the execution time of the same workload on an N-processor system. The paper estimates the speedup of distributed iterative solvers, analyzes the influence of difference communication schemes on the speedup, and compares the estimates with the measurements of real distributed programs.

Keywords:

Distributed computing, speedup, computation-to-communication ratio, sparse systems of linear equations, iterative methods.

References:

Available in pdf and postscript.
Full version of the paper available in pdf.