Medea: a tool for workload characterization of parallel systems

Performance of parallel workload is heavily in uenced by how its own characteristics match the systems Experimental ap proaches are particularly suitable although di cult for study ing the behavior of the programs The raw data collected by monitoring tools on real systems need to be processed for ob taining meaningful and easy to interpret information usable in identifying sources of poor performance MEDEA MEasure ments Description Evaluation and Analysis is a software tool which meets such objectives It has been designed as the nat ural extension of monitors in that it is typically used in strict conjunction with them Various statistical and numerical tech niques are applied to the collected data for deriving compact descriptions of the load i e models able to capture and repro duce its behavior Visualization facilities are an integral part of our tool As applications of MEDEA a kernel and two real life programs i e climate modelling and turbulent ow running in di erent environments are analyzed Hints about possible code This work was supported in part by the the Italian Research Council C N R under grants N PF and N CT and by the Italian M U R S T under the Project Author s current a liation Institute for Software Technology and Parallel Systems University of Vienna Liechtensteinstr A Vienna Austria optimizations most suitable communication policies and data distributions are also discussed