High-Level Modeling Environments for the Dependability Assessment of Dynamic Fault-Tolerant Systems
Tuesday, September 18, 2007
9:00 - 12:30
Offered by:
Antonio Puliafito, Salvatore Distefano - Mobile & Distributed System Laboratory Università di Messina (Italy)
Arndt Bode, Max Walter - Lehrstuhl für Rechnertechnik und Rechnerorganisation, Technische Universität München (Germany)
Marc Bouissou - Électricité de France, EDF (Germany)
Summary
The reliability, availability and safety of fault-tolerant systems
is often estimated by the use of fault trees, reliability block diagrams
or similar Boolean modeling methods. However, with these methods, it is
not possible to include inter-component dependencies and dynamic system
properties like failures with a common cause, failure propagation,
varying failure rates (e.g. for components in standby), shared repair
resources and so on. Thus, Boolean models may lead to over-optimistic results.
On the other hand, the more sophisticated methods based on stochastic
processes (like Markov chains, Petri nets, or models defined using
a stochastic process algebra) do allow for a modeling of these inter-
dependencies and dynamic system properties, but lack user-friendliness:
in comparison, these models are not very intuitive as well as hard to read, modify and
extend and hence difficult to learn and - as a consequence - error-prone.
In this tutorial, we discuss a new generation of methods and tools which
are both simple to use and powerful at the same time. Methods like
dynamic fault trees (DFT), dynamic reliability block diagrams (DRBD) and
their implementation in software tools are based on an automatic generation
of state-based models out of high-level, userfriendly input diagrams.
Real-world case studies serve as a mean to compare the different approaches
and show their benefits in comparison to traditional techniques.
Download tutorial description (pdf)