An Embedded Software Reliability Model with Consideration of Hardware Related Software Failures

As software in an embedded system has taken charge of controlling both software and hardware components, the importance of estimating more accurate reliability for such software has been increased. To estimate the reliability of target software systems, software reliability models are often utilized with software failure data. Since software and hardware are highly co-related and frequently interact with each other in embedded systems, both of them are contributing factors to software failures. Thus, the influence of software and hardware faults on software failures should be taken account for to estimate software reliability. However, many researchers have developed software reliability models assuming that software failures are caused by only software faults, which might lead to inaccurate reliability estimation. In this paper, we suggest two new reliability models considering software and hardware faults as root causes of software failures for embedded software reliability estimation. The proposed models are compared with existing models for validity, and analysis results of the models with real project data are presented. The experimental results show that a Weibull based model, which takes characteristics of hardware degradation into account, has higher fitting-adequacy and superior accuracy for software reliability estimation. Through these results, the proposed model provides more accurate software reliability estimation and helps setting better testing strategies in the earlier phases of the embedded software testing.