Network Topology Abstraction by ALTO

Topology exposure enables a network-aware, optimized resource selection and routing at application level. Application Layer Traffic Optimization (ALTO) is a standardized protocol for such network topology exposure. By using the light-weight and extensible interface network providers can expose information about their internal topology to applications. Promising use cases include Content Delivery Networks~(CDN) and network-aware cloud computing solutions. This paper addresses how to gather and process the information that is needed to offer an ALTO service. We present an architecture that extracts the input map data from information that is already available in network management systems, given their detailed, centralized view of the network. A proof-of-concept prototype demonstrates how data can be obtained and processed using a north-bound interface to a carrier network. Exposing the full network topology of a large network results in large maps that are hardly useful to applications. It also includes information that a network provider might not be willing to expose. Therefore an abstraction is required. We propose using graph transformation algorithms for this reduction of the level-of-detail, and we discuss how to design and evaluate such algorithms. Our numerical results reveal that aggregation can significantly reduce the size of network topologies and the amount of data while still enabling accurate application-level traffic optimization.