Accelerated stress testing to detect probabilistic software failures

Abstract

This paper describes the advantages of accelerated stress testing and addresses the value of accelerated stressing to find probabilistic software failure modes. The value has been widely reported in discovery of traditional hardware-related failure modes, but has not been widely accepted as valuable in finding software errors. This paper provides a business case study in accelerated stressing to find both the traditional hardware failure mechanisms and the elusive probabilistic software failures. This case is presented with a cost benefit analysis including warranty cost avoidance. This business case quantifies the value of accelerated stress testing by demonstrating that in the development phase, accelerated stress testing finds software errors that would have been found later in the product's life cycle after the product is launched and manufactured in high volume, which would have made a detrimental impact to the company's bottom-line.