RNG-based searching and broadcasting over Internet graphs and peer-to-peer computing systems

Abstract In a broadcasting problem, a message is sent from a source to all the other nodes in the network. It can be used for topology updates or to search for a file or information in a peer-to-peer computing system. Blind flooding is a classical mechanism for broadcasting, where each node retransmits received message to all its neighbors. Despite its important advantages, an increase in the number of requests or the size of the routing area produces communication overheads that limit the scalability of blind flooding, especially in networks with dynamic topologies. In this work we propose a new broadcasting and searching scheme over the Internet based on the Relative Neighbourhood Graph (RNG). We extended the definition of RNG in order to use alternative metrics other than geometric. Delay, congestion, random numbers or geographic distances are examples of such metrics that can be used efficiently in RNG based broadcasting. RNG is a sparse connected overlay network which is defined based on local information at each node. This new scheme is compared to the existing Flooding and Rumor Mongering (or Gossip) schemes to evaluate its performance. Our parameterless RNG based scheme guarantees delivery to each node with considerably reduced number of messages with respect to flooding, and has comparable amount of message to Rumor Mongering/Gossip scheme that does not guarantee delivery to each node and also uses parameters whose best value depend on underlying network density.