A hierarchical group of peers in publish/subscribe systems

Abstract

In the P2P (peer-to-peer) type of a topic-based PS (publish/subscribe) (P2PPS) model, each peer can be a subscriber and publisher and directly communicates with other peers in a distributed manner. Messages which have a common publication topic are considered to be related. In our previous studies, the topic vector is proposed to causally deliver related messages published by peers. In addition, messages are assumed to be broadcast to every peer in a system. In scalable systems, it is not easy, maybe impossible to broadcast messages to every peer. In this paper, we consider a two-layered group of multiple peers, which is composed of subgroups, in order to efficiently publish and receive messages. Here, messages are broadcast to every member peer in each subgroup but messages are unicast to subgroups. We propose a GBLC (Group-Based Linear time Causally ordering) protocol for a hierarchical group of peers in a P2P type of topic-based PS model. In the evaluation, we show the number of pairs of unnecessarily ordered messages can be reduced in the GBLC protocol than the linear time (LT) protocol.