Model checking for survivability!

Business and social life have become increasingly dependent on large-scale communication and information systems. A partial or complete breakdown as a consequence of natural disasters or purposeful attacks might have severe impacts. Survivability refers to the ability of a system to recover from such disaster circumstances. Evaluating survivability should therefore be an important part of communication system design. In this paper we take a model checking approach toward assessing survivability. We use the logic CSL to phrase survivability in a precise manner. The system operation is modelled through a labelled CTMC. Model checking algorithms can then decide automatically whether the system is survivable. We illustrate our method by evaluating the survivability of the Google file system using stochastic Petri nets.