Performance models and simulation 

X. Du and X. Zhang     

Chapter 6, High Performance Cluster Computing, Vol. 1, 
Edited by R. Buyya, Prentice Hall, 1999, pp. 135-153.    
 
Abstract
--------

With rapid development and advances of commodity processors
and networking technology,
parallel computing platforms are shifting from expensive and special
purpose massively parallel processing machines
(MPPs) to cheap and general purpose symmetric multiprocessors
(SMPs) and clusters of SMPs/workstations/PCs.
Using commodity SMPs, workstations and/or PCs, we have
many choices to configure a platform for parallel computing.

In this paper, we analyze the performance of a parallel computing
system, which may be a SMP, or a cluster of workstations/PCs using a
parameterized workload characterization. The objective of the work is
to derive a simple and effective analytical method for quickly finding the
nearly optimal platforms for that kind of workload.