Computing system reliability using Markov chain usage models

A Simpler and More Direct Derivation of System Reliability Using Markov Chain Usage Models

“…Pass and fail data are recorded and analyzed for reliability estimation, coverage analysis, or stopping decisions. This form of statistical testing [6,5,10,8,11,14,16,15] supports quantitative certification of software by a statistical protocol. A public domain tool supporting statistical testing (JUMBL: J Usage Model Builder Library developed by UTK SQRL) is freely available [9,1].…”

“…Statistical testing based on a Markov chain usage model, as a rigorous testing method developed by the University of Tennessee Software Quality Research Laboratory (UTK SQRL), has been around for more than two decades [6,5,10,8,11,14,16,15] and successfully applied in a variety of industry and government projects, ranging from medical devices to automotive components to scientific instrumentation, to name a few [3,2,13]. The direct benefit statistical testing provides is a quantitative analysis of the system's quality using empirical data that can be used to demonstrate, document, and certify that the system is fit for its intended use.…”

“…A public domain tool supporting statistical testing (JUMBL: J Usage Model Builder Library developed by UTK SQRL) is freely available [9,1]. This paper reviews the current reliability analysis for statistical testing based on a Markov chain usage model, and specifically how to derive a system end-to-end reliability estimate based on the testing experience given the usage model, as established in [10,12], and proposes a simpler way to compute the expected value of system reliability, not via the system failure probability (or unreliability), but directly. The new derivation results in a simpler closed-form formula with less steps of computation that also renders an intuitive explanation.…”

“…Several reliability models that have been used for statistical testing are presented and compared in [10,12]. They are briefly described below: -The sampling model [7].…”

“…-The arc-based Bayesian model [10,12]. The Miller model is applied to individual arcs of a Markov chain usage model.…”