Accelerated stress testing for both hardware and software

Accelerated stress testing (AST) has been used in electronic, electromechanical, and mechanical systems to achieve robustness with high reliability primarily for hardware. For software products, the reliability program is often conducted separate from any hardware accelerated stress testing. Yet, many systems are consist of concurrent software and hardware issues. In addition, the stress testing processes were primarily adopted by those responsible to develop and manufacture hardware. For example, the stresses usually include temperature extremes, thermal cycles, vibrations, etc. These stresses are effective in accelerating latent hardware defects from degradable, marginal, or intermittent failures to hard failures so that root cause analyses and corrective actions may be made. Although experiments had indicated that software faults and hardware defects are related, the available formulation of the fundamental principles was still based on hardware systems. AST for software and for operating systems have been discussed in [M. Werner, et al., "Improving Customer Satisfaction via Accelerated Stress Testing of General Purpose Computer Operating Systems"] and [M. Werner, et al., 2002], but a fundamental understanding of AST for software is lacking. In order to generalize the fundamentals of accelerated stress testing to address both software and hardware, we need to define accelerated stress testing for software and to address whether they are needed, i.e., whether there are effective methods to achieve high software reliability. The basic reliability concepts categorize systems into different categories according to the presence of defects and faults and whether these weaknesses are explicit enough. The concepts for both hardware and software reliability separate the notion of defects and faults from failures. It further conceptually separates the notion of stressing and the notion of detection. The fundamental concept is that all failures except the explicit ones must be manifested under certain stress conditions. There is then a threshold stress level beyond which a system fails. The cumulative effect of stresses is included by defining time as one type of stress. Both hardware and software systems have marginal weakness, and degradable weakness. The process o- f recovery and repair are also examined for both hardware and software events. The basic reliability principles in accelerated stress testing for both software and hardware systems are combined and explained in this paper. While [M. Werner, et al., "Improving Customer Satisfaction via Accelerated Stress Testing of General Purpose Computer Operating Systems"] and [M. Werner, et al., 2002] also address the needs and advantages of AST for software, an effective software AST program requires efficient tools yet to be developed. The benefits should justify the needed further research and development in this area.