Constraint Based Resilience Analysis

Abstract. In this paper we give an overview of applications of Constraint Programming for IP (Internet Protocol) data networks, and discuss the problem of Resilience Analysis in more detail. In this problem we try to predict the loading of a network in different failure scenarios, without knowing end-to-end flow values throughout the network; the inference is based only on observed link traffic values. The related problem of Traffic Flow Analysis aims to derive a traffic matrix from the observed link traffic data. This is a severely under-constrained problem, we can show that the obtained flow values vary widely in different, feasible solutions. Experimental results indicate that using the same data much more accurate, bounded results can be obtained for Resilience Analysis.

• Type: Conference
• Year: 2006
• Where: CP
• Authors: Helmut Simonis
• Keywords: Artificial Intelligence, CP 2006, Flow Values, Observed Link Traffic, Resilience Analysis