Performance Analytics: Understanding Parallel Applications Using Cluster and Sequence Analysis
Due to the increasing complexity of High Performance Computing (HPC) systems and applications it is necessary to maximize the insight of the performance data extracted from an application execution. This is the mission of the Performance Analytics field. In this chapter, we present three different contributions to this field. First, we demonstrate how it is possible to capture the computation structure of parallel applications at fine grain by using density-based clustering algorithms. Second, we introduce the use of multiple sequence alignment algorithms to asses the quality of a computation structure provided by the cluster analysis. Third, we propose a new clustering algorithm to maximize the quality of the computation structure detected minimizing the user intervention. To demonstrate the usefulness of the different techniques, we also present three use cases.
The work presented in this chapter has been partially founded by IBM, through the IBM-BSC MareIncognito collaboration agreement, the Spanish Ministry of Education under grant BES-2005-7919 and project TIN2007-60625, and the EU/Russia joint project HOPSA.
- 1.Ahn, D.H., Vetter, J.S.: Scalable analysis techniques for microprocessor performance counter metrics. In: ACM/IEEE Conference on Supercomputing (SC), Baltimore (2002)Google Scholar
- 2.Carrington, L., Snavely, A., Gao, X., Wolter, N.: A performance prediction framework for scientific applications. In: 3rd International Conference on Computational Science (ICCS), Saint Petersburg/Melbourne (2003)Google Scholar
- 3.Ester, M., Kriegel, H.P., Sander, J., Xu, X.: A density-based algorithm for discovering clusters in large spatial databases with noise. In: 2nd International Conference on Knowledge Discovery and Data Mining (KDD), Portland (1996)Google Scholar
- 4.Hartigan, J., Wong, M.: Algorithm AS 136: a K-means clustering algorithm. J. R. Stat. Soc. Ser. C (Appl. Stat.) 28, 100–108 (1979)Google Scholar
- 5.Huck, K.A., Malony, A.D.: PerfExplorer: a performance data mining framework for large-scale parallel computing. In: ACM/IEEE Conference on Supercomputing (SC), Seattle (2005)Google Scholar
- 8.Pelleg, D., Moore, A.W.: X-means: extending K-means with efficient estimation of the number of clusters. In: 17th International Conference on Machine Learning (ICML), Stanford (2000)Google Scholar
- 9.Sherwood, T., Perelman, E., Hamerly, G., Calder, B.: Automatically characterizing large scale program behavior. In: 10th International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS), San Jose (2002)Google Scholar
- 11.Vianney, D., Mericas, A., Maron, B., Chen, T., Kunkel, S., Olszewski, B.: CPI analysis on POWER5, Part 2: introducing the CPI breakdown model. http://www-128.ibm.com/developerworks/library/pa-cpipower2