Abstract: The application of existing architecture-based software reliability models is strictly limited to software component transitions satisfying Markov properties.When software has heterogeneous architecture which contains both deterministic and probabilistic transitions，these models intentionally model it as a Markov process.The loss of accuracy of reliability estimation comes as a result.In order to resolve heterogeneous software architectures，a new reliability model is presented based on analysis of component transitions.An algorithm generating Markov chain from heterogeneous software architecture modeled in UML sequence diagram is proposed.A case study shows the new model can yield more accurate reliability estimation result than existing models.

Key words: software reliability, software architecture, heterogeneous architecture, Markov chain, UML sequence diagram

摘要： 现有基于软件体系结构的可靠性模型只适用于符合马尔可夫过程的软件构件转移关系，而当构件间存在由确定性和随机性转移关系组成的异构结构时，仍按照马尔可夫过程对软件体系结构建模，会导致可靠性评估结果不准确。为了解决现有模型不适用于异构结构的问题，在分析构件转移关系基础上，提出了一个新的软件可靠性模型。基于软件UML顺序图，提出将异构软件结构转换为马尔可夫链的算法。通过对一个实例的可靠性评估，验证了新模型能够比现有模型获得更准确的评估结果。

关键词: 软件可靠性, 软件体系结构, 异构软件体系结构, 马尔可夫链, UML顺序图