Petri Net-based Reliability Research For Component-based Software System

Component-based technology has become the focus of software engineering research and the key of developing large scale software systems. In order to improve the reliability of the software system, it has become a crucial step to analyze the reliability of the system using its architecture—its inner components and interfaces construction information—in the process of software development.We propose a Petri net-based method for component-based software reliability analysis, using Petri net to simulate and analyze component-based software in a dynamic mode. First, according to the software architecture we employ Petri net to model the system; then we take advantage of the dynamic analysis technique of the Petri net to analyze the construction characteristics and reliability factors of the system in a comprehensive way; finally we can calculate the total reliability of the system using reliability data on components and interfaces which we have already known. This model uses stochastic time of the stochastic Petri net to describe the transition running time of the component-based software, analyzing time influence on the software reliability in the process of Petri net modeling. Varying from many other models which concentrate only on execution paths or ransition probility between components, this model not only gives attention to static structure features of the system, but also considers dynamic running characteristics of the system, solving loop-path, running time, interface calling problems in component-based software reliability analysis in an elegant manner, making the result of reliability analysis more accurate. In this process the system designer can apply sensitivity analysis to the components and interfaces that construct the system. Sensitivity analysis enabls him to reconstruct the system, developing or choosing more reliable components to replace old ones so as to upgrade the performance of the system.The Petri net-based analysis method we propose can evaluate all sorts of reliability parameters of the component-based software effectively and accurately. It can utilize other appropriate component-based software reliability calculation models to accommodate other special applications. Not only can it predict the reliability of component-based software in the early stage of software development, but also can evaluate the reliability of component-based software in the developing, testing, functioning and maintaining phase. It is a useful method to guide the analysis and design of software architecture and to enhance the reliability of component-based software systems.