Sharing the cost of backbone networks: Simplicity vs. precision

Internet backbone operators face a trade-off in quantifying the costs that their customers inflict on their infrastructure since the precision of these methods depends on the resources dedicated to traffic monitoring. Operators prefer simple and straightforward monitoring schemes, which in turn raises the question of: "what is the price of simplicity in monitoring and cost computation?" We address this questions by quantifying the costs of customers with real-world data. Our four-week long dataset describes a large operator's customer traffic patterns over 401 geographically distributed network links. We study the differences of common cost sharing policies compared to an absolutely accurate and fair policy. The cost discrepancy between different methods exceeds 25% in 71% of the examined cases. We reveal the root cause of these discrepancies by analyzing the customer traffic patterns. Moreover, we quantify the impact of the geographically diverse costs of network links. Our results reveal that simplicity comes at a rather high cost in terms of decoupling between computed and real costs.