Saturday, October 6, 2012

"A Methodological Framework for Software Safety in Safety Critical Computer Systems"

The Journal of Computer Science is frequently overlooked in the Embedded Space for the solutions to many problems, for example the September issue covered topics as diverse as, Speed Control of Switched Reluctance Motor Using New Hybrid Particle Swarm Optimization for the hardware types and among us, and Fuzzy Cost Enabled Cluster Based Multipath Routing Algorithm for Mobile Ad-Hoc Networks for those putting together the latest sensor net.

Of particularly interest to me is A Methodological Framework for Software Safety in Safety Critical Computer Systems [PDF] by P. V. Srinivas Acharyulu and P. Seetharamaiah. These authors have put together one of the best introductions to issues related to the safety of systems controlled by software, from defining the terms to building a model system out of a real model train set to demonstrate the techniques described.

Abstract

"Software safety must deal with the principles of safety management, safety engineering and software engineering for developing safety-critical computer systems, with the target of making the system safe, risk-free and fail-safe in addition to provide a clarified differentaition for assessing and evaluating the risk, with the principles of software risk management. Problem statement: Prevailing software quality models, standards were not subsisting in adequately addressing the software safety issues for real-time safety-critical embedded systems. At present no standard framework does exist addressing the safety management and safety engineering priniciples for the development of software safety in safety-critical computer systems. Approach: In this study we propose a methodological framework involving safety management practices, safety engineering practices and software development life cycle phases for the development of software safety. In this framework we make use of the safety management practices such as planning, defining priniciples, fixing responsibilities, creteria and targets, risk assessment, design for safety, formulating safety requirements and integrating skills and techniques to address safety issues early with a vision for assurance and so on. In this framework we have also analysed integration of applicability of generic industrial heirarchy and software development heirarchy, with derived cyclical review involving safety professionals generating a nodal point for software safety. Results: This framework is applied to safety-critical software based laboratory prototype Railroad Crossing Control System (RCCS) with a limited complexity. The results have shown that all critical operations were safe and risk free. Conclusion: The development of software based on the proposed framework for RCCS have shown a clarified and improved safety-critical operations of the overall system peformance."

Journal of Computer Science
DOI: 10.3844/jcssp.2012.1564.1575
Volume 8, Issue 9
Pages 1564-1575

Do keep in mind it is an introduction, a real Rail Road Crossing Control System (RCCS) would be made from two or more systems running parallel, using different processors, different hardware developed by different teams in different languages such C and ADA. The real fun part is developing a system that must run for a few decades and not the eighteen month life span for many components today.