Publish/Subscribe Techniques For P2P Networks

Introduction

A publish/subscribe networking system is one in which the nodes can serve the role of a publisher or a subscriber to publish data or subscribe for data of interest, respectively. The publish/subscribe model differs from other request/response models in that a query of the former model is submitted and stored in advance, for which the result may not yet exist but the query subscriber expects to be notified if and when the result later becomes available. The publish/subscribe model is thus suitable for search applications where queries await future information, as opposed to the traditional applications where the information to be searched must pre-exist.

Enabling publish/subscribe services in peer-to-peer (P2P) networks is a topic that has received a lot of attention in recent years. As P2P can be adopted for distributed networking as an effective way to share resources, minimize server costs, and promote boundary-crossing collaborations, a publish/subscribe functionality should be useful to these networks. For example, a monitoring operator in a P2P-based geographical observation network (Teranishi et al., 2008) will be able to subscribe a query to receive alerts of fire occurrences so that necessary rescue efforts can be dispatched quickly; or, in a P2P-based scientific information sharing network (Shalaby & Zinky, 2007), a subscriber will be notified when new scientific information is published.

Usually, a publisher node does not know who is interested in its data, and, vice versa, a subscriber node does not know where in the network its data of interest is available. Thus, a challenging problem is to design mechanisms for the subscribers and publishers to find each other quickly and efficiently. A simple way is to broadcast each query to all the nodes in the network or to employ a centralized index of all the queries subscribed and information published. This mechanism is neither efficient nor scalable if applied to a large-scale network.

Consequently, a variety of distributed publish/subscribe mechanisms have been proposed. They follow two main approaches: gossip-based and structure-based. The first approach is designed for any unstructured networks, in which the subscriber nodes and publisher nodes find each other via exchanges of information using the existing peer links, typically based on some form of randomization. The other approach organizes the nodes into some overlay structure and develops publish/subscribe methods on top of it. Examples of such an overlay are those based on Distributed Hash Tables (e.g., CAN (Ratnasamy et al., 2001), Chord (Stoica et al., 2001). The gossip-based approach’s advantage is its applicability to any unstructured network, while the structure-based approach is favored for better efficiency.

This chapter provides a survey on the publish/subscribe techniques for P2P networks. First, we will provide some necessary preliminaries. We then discuss several representative techniques in each of the following categories: structure-based, gossip-based, and a hybrid of these two. We conclude the chapter with some remarks.