PERFORMANCE MODELS FOR SCALABLE CLUSTER COMPUTING

In this paper, we make a thorough study of scalable parallel computing performance models, point out some of the weaknesses of existing perf ormance models, and discuss how one can improve and generalize them. W e start with analyzing and accurately defining several kinds of serial and parallel execution times, discuss parallel speedup models, and pr esent a generalized speedup model. We propose a formal definition of s calability, and discuss scalability of cluster systems from three aspe cts. As an extension to the isoefficiency metric and isospeed metric, we propose and define scalability of parallel algorithm-architecture c ombinations based on the fixed ratio of computation time to communicat ion overhead, analyze some relevant theoretical perspectives, and comp are scalability with speedup.