Testing Self-Similar Networks

Abstract

A hard problem in network testing is verifying the correctness of a class of networks, as well as the actual networks under test. In practice, at most a few networks (sometimes only one) are actually tested. Thus an important question is how to select one or more networks that are sufficiently representative to apply the results to a class of networks. We present a model-based technique for selecting a representative network. The central theorem establishes that the representative network displays any faults present in any network of the class. This paper introduces the concept of "self-similarity," which is used to select the network, and presents the results of an experiment in testing one class of networks. © 2006.