Modeling silent failures in telecommunications systems

Abstract

A silent failure is a condition in a telecommunications or other system which, when it occurs, remains undetected by normal fault detection methods. With the steady state Markov models that are usually used to predict telecommunications system reliability, estimates of downtime of systems with silent failures may be substantially higher than actual system downtime. This is due to the fact that a system with silent failures frequently comes nowhere close to approaching steady state during the system's finite life. This paper proposes a modification to the standard steady state Markov reliability models. The proposed modification involves the addition of a state transition effectively representing complete replacement of a system under study. With the modified model, this transition occurs at a rate 2/T, where T is system or study life. The paper includes examples and theorems that demonstrate that the method produces accurate results in a wide range of circumstances.