News Archives

Performance Prediction of Software Performance: A Model Interoperability Approach

February 19, 2009

  • Date: Thursday, Feburary 19th, 2009 
  • Time: 11 am — 12:15 pm 
  • Place: ME 218

Connie U. Smith
L&S Computer Technology, Inc

Abstract: Performance, both responsiveness and scalability, is an important quality of today’s software. Yet many software systems cannot be used as they are initially implemented due to performance problems. These performance failures can translate into significant costs due to damaged customer relations, lost income, and time and budget overruns to correct the problem. Our experience is that performance problems are most often due to fundamental architectural or design problems rather than inefficient coding. Thus, performance problems are introduced early in the development process but are typically not discovered until late, when they are more difficult and costly to fix.

The talk first introduces the Software Performance Engineering (SPE) approach for predicting performance during the early stages of development, before the architecture is fully determined. Next, the software and system performance model technology and data requirements are explained. Then an overview of the foundations of the software performance model interoperability approach is presented. It presents two performance model interchange formats: S-PMIF and PMIF, and proof of concept results of model interoperability in the SPE process. We briefly cover recent extensions to the model interoperability approach, and conclude with a discussion of future work.

Bio: Connie U. Smith, Ph.D. is a principal consultant of the Performance Engineering Services Division of L&S Computer Technology, Inc. She received a BA in mathematics from the University of Colorado and MA and Ph.D. degrees in computer science from the University of Texas at Austin. She is the author of Performance Engineering of Software Systems published in 1990 by Addison-Wesley, Performance Solutions: A Practical Guide to Creating Responsive, Scalable Software published in 2002 in Addison-Wesley’s Object Technology Series, and approximately 100 scientific papers. She is the creator of the SPE-ED™ performance engineering tool and collaborated on developing several performance model interchange formats. She has over 25 years of experience in the practice, research and development of the SPE performance prediction techniques, computer performance modeling and evaluation, performance patterns and antipatterns, tool interoperability, and tool development. Dr. Smith received the Computer Measurement Group’s prestigious AA Michelson lifetime achievement award for technical excellence and professional contributions for her SPE work.

She frequently serves on conference and program committees, including founding and chairing the First International Workshop on Software and Performance (WOSP) in 1998, serving on Conference and Program Committees for subsequent WOSP conferences, and currently leads the WOSP Advisory Committee. She served as an officer of ACM SIGMETRICS for 10 years, is a past ACM National Lecturer, is an active member of the Computer Measurement Group and the Quantitative Evaluation of Systems (QEST) conferences. She was previously a faculty member of the computer science department at Duke University. Since then she has been with L&S Computer Technology specializing in the development and support of the software performance engineering tool, SPE-ED™, applying performance prediction techniques to software, teaching SPE seminars, and research and writing on SPE. Dr. Smith can be reached by email at {mailto address=”cusmith@spe-ed.com” encode=”javascript” text=”cusmith@spe-ed.com” } . A list of publications may be found at http://www.spe-ed.com.