Improving Performance Analysis Using Resource Management Information
- 308 Downloads
In this paper we present Clane, a performance analysis environment for clusters. It uses a novel approach combining resource management and monitoring data to provide reliable information for cluster users and administrators in application and system performance analysis. Clane uses the XML standard to represent its internal information base, providing more flexibility in data manipulation and simplicity to extend the environment with other analysis tools. The environment is composed by an Information Server, which stores performance information provided by the monitoring system and user events dispatched through the resource management system, and an Analysis Tool to present the combined information and events using statistics, graphs and diagrams. It also enables performance comparisons among distinct executions of the same application in the cluster.
KeywordsMonitoring System Resource Management System Application Execution Cluster User Monitoring Session
Unable to display preview. Download preview PDF.
- 1.Buyya, R.: High Performance Cluster Computing: Architectures and Systems, vol. 1. Prentice-Hall, Englewood Cliffs (1999)Google Scholar
- 2.Henderson, R.L., et al.: Portable batch system: Requirement specification. Technical report, NASA Ames Research Center (1995)Google Scholar
- 3.Keller, A., Reinefeld, A.: Anatomy of a Resource Management System for HPC Clusters. Annual Review of Scalable Computing 3 (2001)Google Scholar
- 4.Baker, M.: Cluster computing white paper (2000)Google Scholar
- 5.C.L.S., et al.: Myrinet – a gigabit-per-second local-area network. IEEE Micro 15 (1995)Google Scholar
- 6.IEEE standart 1596-1992 New York: IEEE: IEEE Standart for Scalable Coherent Interface, SCI (1993)Google Scholar
- 7.Goodwin, M., et al.: Performance Co-Pilot User’s and Administrator’s Guide. Silicon Graphics, Inc. (1999)Google Scholar
- 9.Center, C.R.: CPAD-PUCRS/HP (2003), http://www.cpad.pucrs.br
- 10.Netto, M.A., Rose, C.D.: Crono: A configurable management system for linux clusters. In: Proceedings of the 3rd LCI International Conference on Linux Clusters: The HPC Revolution 2002 (LCI 2002), St. Petersburg, Florida (2002)Google Scholar
- 11.Ferreto, T., Rose, C.D., DeRose, L.: Rvision: An open and high configurable tool for cluster monitoring. In: Proceedings of the 2nd IEEE/ACM International Symposium on Cluter Computing and the Grid (CCGrid 2002), Berlin, Germany, pp. 75–82 (2002)Google Scholar
- 12.Mills, D.L.: Internet time synchronization: the network time protocol. IEEE Trans. Communications, 1482–1493 (1991)Google Scholar