Content networks: distributed routing decisions in presence of repeated queries

Content networks are overlay networks, enabling access to distributed contents on centralized servers or individual computers. Since the flooding-based routing scheme features poor scalability, we present a modification, which reduces the total network traffic while retaining the original efficiency. In choosy routing, each node, while passing an answer, remembers where it came from. Subsequently repeated queries about the same content are forwarded only to one neighbor. This way, the network learns effective routes. The simulations on several topology types have shown the expected behavior, with up to three-fold reduction in the overall query traffic.