SanFrancisco performance: a case study in performance of large-scale Java applications

Authors:
R. Christ;S. L. Halter;K. Lynne;S. Meizer;S. J. Munroe;M. Pasch
Affiliations:
IBM Software Group, Rochester Laboratory, 3605 Highway 52 North, Rochester, Minnesota;IBM Software Group, Rochester Laboratory, 3605 Highway 52 North, Rochester, Minnesota;IBM Software Group, Rochester Laboratory, 3605 Highway 52 North, Rochester, Minnesota;IBM Software Group, Rochester Laboratory, 3605 Highway 52 North, Rochester, Minnesota;IBM Software Group, Rochester Laboratory, 3605 Highway 52 North, Rochester, Minnesota;IBM Software Group, Rochester Laboratory, 3605 Highway 52 North, Rochester, Minnesota
Venue:
IBM Systems Journal
Year:
2000

Citing 2
Cited 4

Thin locks: featherweight synchronization for Java

PLDI '98 Proceedings of the ACM SIGPLAN 1998 conference on Programming language design and implementation
Java server benchmarks

IBM Systems Journal

JaViz: a client/server Java profiling tool

IBM Systems Journal
Formalising Java RMI with explicit code mobility

OOPSLA '05 Proceedings of the 20th annual ACM SIGPLAN conference on Object-oriented programming, systems, languages, and applications
Formalising Java RMI with explicit code mobility

Theoretical Computer Science
Optimising Java RMI programs by communication restructuring

Proceedings of the ACM/IFIP/USENIX 2003 International Conference on Middleware

Quantified Score

Hi-index	0.00

Visualization

Abstract

This is a study of the performance progress of the SanFranciscoTM project from the time the decision was made to base it on the JavaTM language up to the time SanFrancisco applications were successfully deployed in the marketplace--from February 1997 until late 1999. We document the challenges, decisions, and technologies that were encountered during the three-year development period that saw performance improve by orders of magnitude. Key areas that allowed us to achieve this improvement were intelligent object caching, improved object access strategies, use of commands (function shipping), efficient mapping of objects to the underlying database, appropriate usage of Java, programmer education, and acquiring (or building) needed tools. We also discuss several areas where challenges remain and more progress is needed.