Towards a statistical characterization of the competitiveness of oblivious routing

Oblivious routing asks for a static routing that serves arbitrary user demands with minimal performance penalty. Performance is measured in terms of the competitive ratio, the proportion of the maximum congestion to the best possible congestion. In this paper, we take the first steps towards extending this worst-case characterization to a more revealing statistical one. We define new performance metrics and we present numerical evaluations showing that, in statistical terms, oblivious routing is not as competitive as the worst-case performance characterizations would suggest.