Quantitative Aspects of Designing and Validating Dependable Computing Systems -- Calculations, Measurements, and Simulations

Summary

Increasingly complex computing systems are used for a variety of applications, from world-wide networks to massively parallel computing, from transaction processing to automation of safety-critical systems. The usefulness and success of these applications depends on hardware and software dependability. Achieving them can be rather difficult and may require highest efforts because system complexity is the origin of many design faults, especially in the software, and the required system performance often is in conflict with the redundancy needed for fault tolerance.

Consequently, a careful system design is required. Besides design tools like silicon compilers and CASE, special methods for the prediction and evaluation of system dependability play an important role. In contrast to some former reliabilility calculations dependability evaluation must be able to cope with complex system behaviour. One cannot expect that straight-forward methods are able to cover an extremely large state space or express the behavior of highly interacting subsystems with sufficient accuracy.

This seminar was concentrated on new methods to quantify the dependability of complex computing systems. The usefulness of design and validation techniques for various types of highly dependable systems was assessed and discussed. It clearly turned out that a sufficient dependability evaluation cannot be achieved by just a single method. Both the design process and the system operation must be accompanied by appropriate dependability quantification techniques from the very beginning of the conceptual phase throughout the system lifetime. Moreover, particular efforts are necessary for design fault quantification as well as field data collection and interpretation – appropriate methods must take into account that even extremely rare fault events have an impact on the quantification of highest dependability.