Estimating Software System Reliability: An Architecture-Based Approach

To ensure high reliability for large and complex software systems, fault tolerance mechanisms have been widely used in the system design processes. However, less attention has been paid to the analysis of the system reliability improvement concerning fault tolerance. Therefore, they might underestimate the system reliability. To remedy this, we extend architecture based reliability estimation to take account of failure recovery mechanisms. This has two elements. Firstly, from the view of correctness of components´ I/O, we give an analytical model for the architecture-centric system to detailly illustrate the invocation among components. Second we leverage Discrete-Time Markov Chain (DTMC) to calculate the system reliability according to the analytical model. The results help to locate the crucial components that need to be combined with robust fault tolerance mechanisms to improve the system reliability efficiently. Case study shows how our approach can produce an accurate estimation of system reliability.