SPICE: Scalable P2P implicit group messaging

Implicit group messaging (IGM) is a decoupled messaging paradigm for connecting content publishers and consumers over the Internet. Unlike traditional multicast or publish/subscribe messaging, IGM delivers content to “implicit groups” of consumers with characteristics specified by the publisher at the time of publication. IGM systems must support thousands of users and an infinite number of implicit groups formed on demand as messages are published. These groups may be messaged repeatedly or once only, with group sizes scaling from no members to the entire network. Load distribution is a key problem of such systems. This paper broadens our earlier work [D. Cutting, B. Landfeldt, A. Quigley, Implicit group messaging over peer-to-peer networks, in: A. Montresor, A. Wierzbicki, N. Shahmehri (Eds.), Sixth IEEE International Conference on Peer-to-Peer Computing (P2P2006), IEEE Computer Society, Cambridge, United Kingdom, September 2006, pp. 125–132.] in three ways: we provide a formal specification of implicit groups and implicit group messaging; we introduce a comprehensive framework for analysing the efficiency and fairness of generic IGM implementations; and our distributed structured peer-to-peer IGM model, Spice, is augmented with adaptive load distribution techniques. Through detailed simulation and analysis using Zipfian data sources we demonstrate these techniques are capable of very fairly distributing incoming and outgoing loads over peers irrespective of the scale of implicit groups or frequency of messages.