Is cross-domain fault localization feasible?

Abstract

Troubleshooting network faults is challenging. It is even more difficult to diagnose cross-domain issues without complete knowledge of observations and topology from neighboring network domains. For both competitive and security reasons, domain managers hesitate to share observations even when doing so may significantly ease fault localization. In this paper, we present the first comprehensive evaluation of the feasibility of cross-domain fault localization. Leveraging a recent graph-digest based formulation of the problem, we have developed a set of practical metrics and performed extensive experiments to evaluate two key questions: Does cross-domain fault localization offer the kinds of benefits warranting further research? Can it provide deployable and acceptable privacy protection with manageable complexity? We evaluate both provider-customer and peering relationship scenarios between small, medium, and large domains. Our results show that crossdomain fault localization is effective using the graph-digest approach and that sensitive network connectivity properties can be concealed from other domains.