Making historical connections: Building Application Layer Traffic Optimization (ALTO) network and cost maps from public broadband data

The Application Layer Traffic Optimization (ALTO) protocol allows network service providers to make available a pair of maps to applications such that the applications can intelligently (compared to randomly) connect to a desired resource. The network map aggregates the service provider network into provider defined identifiers (PID) and the cost map provides a pair-wise link cost between each PID. Clearly, a service provider has an authoritative view of its network and is able to provide an ALTO server that distributes such maps. However, ALTO also envisions third-parties as being able to provide such maps. In this paper, we demonstrate how a third-party ALTO server can provide maps by mining public information. Specifically, we build our maps from the United States Federal Communications Commission public broadband data set, which contains an expressive (multi-tier wireline broadband measurements) and rich (measurements for specific application uses) dataset. In all, we examined over 1 billion records spread over 90 GBytes as part of our analysis. We borrow concepts from financial engineering and social network analysis to show how network topology and cost maps can be created, and furthermore, how peer-to-peer systems can insulate themselves from going dark by choosing supernodes effectively from mining historical data.