Specific problems of critical systems

Technology systems, presenting important stakes in dependability, are now designed, manufactured and operated in varying ways by different sectors (nuclear, aviation, rail, automotive industry), but still fundamentally represent a combination of certain principles such as  

Applying these principles, adapted to computer technology, to the challenges of dependability, to economic constraints and regulations, to histoiy and to the experience of each professional sector, forms a state of the art enabUng the design, build, and operation of a safe systems, i.e. with a level of risk tolerated by society. 

Given this state of the art, much of the risk of a safe system does not come from random failures, but from systemic failiues , affecting an entire system, such as common cause failures simultaneously impacting all the hardware of a redundant system, systematic software failures due to design errors, failures associated with a domino effect, failures due to bad supervision or natural disasters, simultaneously affecting initiating events and barriers, etc. 

Systemic risk can thus be defined as the potential for systemic failure. Systemic approaches have been, or are being, implemented to control complexity, particularly through the creation of global installation models designed to smdy important properties such as safety or availability. These quantitative RAMS (Reliabilty, Availability, Maintainability and Safety) models, based on probabilistic approaches, are fairly well accepted for their representation of pure hardware elements (random failures) and human factors. They are much less so for their representation of deterministic aspects of computer systems (systemic or systematic failures). 

However, programmed techniques do not only have benefits. Other than a short shelf-life and rapid aging of many programmed equipment, the main difficulties come from the introduction of new failure modes and aging and from a significant 

---