Model-Driven Design of Stable Software Systems: A Petri Net Based Approach

摘要

In this paper, stability of event-driven software systems is studied in terms of its operational failures due to poor design and/or implementation flaws. The primary objective of our work is to develop a usable model driven design methodology for stable software system design, build a Petri net (PN) based application tool and illustrate its use through a simple case study. PN based techniques are employed to analyze factors that affect software stability, by utilizing PN properties such as boundedness, reachability and reversibility. An analysis tool, written using MATLAB software, is developed and utilized to identify sources of these problems. The analysis helps in delineating those points (transitions) in the software system that need to be monitored to prevent unstable operation. These transitions can then be supplemented by guard functions so as to alleviate problems arising from irreversibility, unboundedness, and unreachability. The tool can also be used to: (i) Detect partially disconnected and completely disconnected subnets to verify reachability and irreversibility, (ii) Detect unbounded places and states in the PN model to verify unboundedness, (iii) Detect minimal empty siphons (potential deadlocks) to check for deadlocks, another source of irreversibility.