Efficient content-based routing with network topology inference

Content-based publish/subscribe has gained high popularity for large-scale dissemination of dynamic content. Yet it is highly challenging to enable communication-efficient dissemination of content in such systems, especially in the absence of a broker infrastructure. This paper presents a novel approach that exploits the knowledge of event traffic, user subscriptions and topology of the underlying physical network to perform efficient routing in a publish/subscribe system. In particular, mechanisms are developed to discover the underlay topology among subscribers and publishers in a distributed manner. The information of the topology and the proximity between the subscribers to receive similar events is then used to construct a routing overlay with low communication cost. Our evaluations show that for internet-like topologies the proposed inference mechanisms are capable of modeling an underlay in an efficient and accurate manner. Furthermore, the approach yields a significant reduction in routing cost in comparison to the state of the art.