See 2010

Evaluating Repair Strategies for a Water-Treatment Facility using Arcade


Stephan Roolvink


18 June 2010






The performance and dependability of critical infrastructures, such as water- treatment facilities is essential. These critical infrastructures are an intricate part of our daily lives as everybody needs clean drinking water, gas to cook food and electricity to cool their food. Without these thing society would grind to a halt. With the control networks of these infrastructures becoming more interconnected with the Internet they are becoming more susceptible to malicious attacks. Luckily, governments from around the globe have recently increased their focus on these infrastructures in an effort to assure a stable society and a healthy economy.

Using various performance and dependability measures we analyse a simplified model of a water treatment facility. We have defined a new measure, survivability which allows us to analyse the recovery of the model Given the Occurrence Of a Disaster. Building on the existing architectural framework Arcade, a model is derived in XML format and then automatically mapped to the model checker PRISM. Using the stochastic model checking capabilities that PRISM offers, we compare different repair strategies, with respect to their costs, system reliability, availability and survivability.

From the case study we conclude that dedicated repair is a fast but very expensive repair strategy and Faster Failure First with one repair crew is slow but in expensive. However, the other repair strategies are fast and inexpensive and depending on the need of the company one or the other will be better. In general we conclude that, when using this method, a company can gain useful insight into the best choice for a repair strategy with respect to performance, dependability and costs. Using our method new measures can be easily expressed and efficiently computed. For future work we want to apply the approach to larger infrastructures and include service degradation cost to our models.