Error analysis and reliability metrics for software in safety critical systems

Abstract

Safety critical systems often rely on hardware only solutions because there exist well-established, consistent, repeatable processes for measuring hardware reliability. But as systems become more complex, incorporating software into safety critical systems is an attractive option. Previously we demonstrated a method to specify provably correct software modules for use in hardware/software safety critical systems. However, there is still a need to develop techniques to demonstrate software reliabilty using metrics comparable to those available for hardware. The method that we present here demonstrates a testing and error recording process that can support a software reliability metric comparable to hardware reliability metrics in strength and repeatability. We also describe a proposed set of standard benchmarks which can be used to help evaluate software reliability measurement processes.