Complex Software Systems: Two Non-Software Reliability Concerns
Ephraim Suhir*, Zoran Stamenkovic, Artem Ponomarev and Eugene Suhir
Abstract
The objective of this write-up is to draw attention of reliability engineers working in the field of highly popular today complex software systems (CSS), like, e.g., artificial intelligence (AI), large language models (LLM) and robotics engineering (RE), that when it comes to assuring success and safety of such systems, non-software reliability, such as the reliability of the systems’ hardware and human performance, are equally, if not even more, important. Two more or less typical problems are addressed here as suitable examples: 1) the hardware reliability of PN junctions employed in semiconductor electronics for almost a century and 2) the ergonomics-engineering human-in-the-loop (HITL) role in critical undertakings, when the hardware reliability of the electronic-and-photonic materials-and-structures and human performance contribute jointly to the success and safety of an undertaking of importance. The emphasis is on the need for quantifying, preferably on the probabilistic basis, the possible role, significance, attributes, and challenges, associated with the application of the “old-fashioned” analytical (“mathematical”) predictive modeling (APM) technique, as well as on the experimental foundation of this technique – failure oriented accelerated testing (FOAT).
