The PODS diversity experiment

A high integrity system typically has a number of redundant components operating in parallel to reduce the probability of a system failure. If the component failures were random, then the probability of several components failing simultaneously would be much smaller than the failure probability of any single component. However, should the components contain common design flaws, then more than one component could fail simultaneously due to a common cause (a common mode failure). This would increase the probability of a system failure. For a computer-based system where the same software component is being run in each processor, any software fault is a potential cause of common mode failure. One method of reducing common software faults is to use diverse software in each processor (n-version programming). This paper discusses how the use of software diversity raises a number of issues. A diversity experiment is described