Content Location and Distribution in Converged Overlay Networks

A major challenge for organizations and application service providers (ASP) is to provide high quality network services to geographically dispersed consumers at a reasonable cost. Such providers employ content delivery networks (CDNs) and overlay networks to bring content and applications closer to their service consumers with better quality. Overlay networks architecture should support high-performance and highscalability at a low cost. For that end, in addition to the traditional unicast communication, multicast methodologies can be used to deliver content from regional servers to end users. Another important architectural problem is the efficient allocation of objects to servers to minimize storage and distribution costs. In this work, we suggest a novel hybrid multicast/unicast based architecture and address the optimal allocation and replication of objects. Our model network includes application servers which are potential storage points connected in the overlay network and consumers which are served using multicast and/or unicast traffic. General costs are associated with distribution (download) traffic as well as the storage of objects in the servers. An optimal object allocation algorithm for tree networks is presented with computational complexity of O(N2). The algorithm automatically selects, for each user, between multicast and unicast distribution. An approximation algorithm for general networks is also suggested. The model and algorithms can be easily extended to the cases where content is updated from multiple locations.