Multi-Criteria Decision Aggregation and Ranking Approaches for Combining Multiple Information Strategies to Decide Which Agents should be Replicated INO CARE III “Towards Fault-Tolerant Cooperative Air Traffic Management” Project Research Report 9 November 2007 Jean-Pierre Briot, Alessandro Almeida, Isabelle Alvarez, Zahia Guessoum, Olvier Marin, Jean-Fran¸ cois Perrot Laboratoire d’Informatique de Paris 6 (LIP6), Universit´e Paris 6 - CNRS Case 169, 4 place Jussieu, 75252 Paris Cedex 05, France {Jean-Pierre.Briot, Alessandro.Luna-Almeida, Isabelle.Alvarez Zahia.Guessoum, Jean-Francois.Perrot}@lip6.fr Abstract This work focuses on the engineering of software replication techniques for distributed cooperative applications designed as multi-agent systems. Such applications are often very dynamic: e.g., new agents can join or leave, they can change roles or strategies. Also, the relative importances of agents may evolve during the course of computation and cooperation, as opposed to tra- ditional static approaches of replication, e.g., for data bases, where critical servers may be identified at design time. Thus, we need to dynamically and automatically identify the most critical agents and to adapt their replication strategies (e.g., active or passive, number of replicas), in order to maximize their reliability and their availability. An important issue is then: what kind of information could be used to estimate which agents are the most critical agents ? In this report, we quickly summarize our approach and prototype architecture for adaptive replication. In a companion report [8], we discuss various kinds of metrics to estimate criticality of agents: static dependences, dynamic dependences, roles, norms, and plans. In this technical report, we discuss the possibility of combining among these various informations along a multi-criteria decision approach. At first, we discuss aggregation methods. They are effective, but leave the designer with the responsability of setting the weights for the aggregation. We then discuss a ranking approach as a way to address this limitation. 1