Techniques In Scalable And Effective Parallel Performance Analysis

Performance analysis tools are essential to the maintenance of efficient parallel execution of scientific applications. As scientific applications are executed on larger and larger parallel supercomputers, it is clear that performance tools must employ more advanced techniques to keep up with the increasing data volume and complexity of the performance information generated by these applications as a result of scaling. In this thesis, we investigate the useful techniques in four main thrusts to address various aspects of this problem. First, we study how some traditional performance analysis idioms can break down in the face of data from large processor counts and demonstrate techniques and tools that restore scalability. Second, we investigate how the volume of performance data generated can be reduced while keeping the captured information relevant for analysis and performance problem detection. Third, we investigate the powerful new performance analysis idioms enabled by live access to performance information streams from a running parallel application. Fourth, we demonstrate how repeated performance hypothesis testing can be conducted, via simulation techniques, scalably and with significantly reduced resource consumption. In addition, we explore the benefits of performance tool integration to the propagation and synergy of scalable performance analysis techniques in different tools.unpublishednot peer reviewe