System test cost modelling based on event rate analysis

Abstract

Unlike IC and board level test, system complexity generally limits the number of methods available to support cost-optimised system test strategy development. This paper describes a parameterised model of system behaviour during both production testing and initial field run-time. The model represents the occurrence rate of error and failure events under test and application workloads and the resulting parameters directly characterise system test effectiveness. These event rate models are fitted to actual data and incorporated into a cost function which calculates overall "cost of test" in relation to key variables. The approach is applicable to both hardware and software related events and promotes a customer view of system quality.