Combined Perfromance and Dependability Evaluation with Conjoint Simulation

Abstract

The paper presents an approach for combining different modeling techniques to facilitate performance and dependability evaluation of parallel computing systems. The approach is called Conjoint Simulation and is based on process-based and object-oriented paradigms in combination with Petri net modeling. The overall system is split up into a performance- and a dependability-related part which are represented and analyzed using different modeling techniques. Conjoint Simulation facilitates model representation, development, and maintenance, and it also enhances model evaluation through the conjunction of the different model parts which are independently developed and modified.


Authors' Biographies:

Axel Hein received the Diploma in Computer Science from the University of Erlangen-N|rnberg and joined the Institute for Computer Science III in 1991. His current research interests include analytical and simulation-based modeling techniques for the performability analysis of fault-tolerant parallel computer systems.

Kumar Goswami received his PhD from the University of Illinois. He is currently working in the performance engineering group at Tandem Computers. His interests include performance and architecture of I/O subsystems, multimedia systems, video on demand as well as the development of performance analysis tools and enterprise systems.