On the performance of two topologies for a distributed broker hierarchy of publish-subscribe middleware

Abstract

Distributing the right information at the right time is an apparent need of the fast moving globe. Publishsubscribe communication paradigm plays a major role in the systems built to accomplish this need. It has been adopted in variety of today’s business domains such as mobile communication, database integration, road traffic visualization etc. With large-scaled, distributed and heterogeneous nature of these systems, there is a high demand for an efficient, scalable and interoperable messaging middleware which is capable of handling the increasing load of messages. As discussed in earlier works, network of messaging brokers that collaboratively act as a single entity provides a scalable architecture for such publish-subscribe middleware. This paper describes two algorithms named tree and cluster that we designed which are used as two approaches for topology of the broker network and evaluates the performance of the distributed publish-subscribe middleware with respect to those two topologies.