Modelling Peer-to-Peer Data Networks Under Complex System Theory

A Peer-to-peer Data Network (PDN) is an open and evolving society of peer nodes that assemble into a network to share their data for mutual benefit. PDNs are enabled by distributed query processing. We argue that with a self-organizing, dynamic, and large-scale architecture and nodes that inherit extensive autonomy from their human users, this new generation of distributed database systems should be upgraded from the domain of traditional distributed computing systems to the realm of natural complex systems (such as social networks). In this way, PDN is studied among its peers within a modelling framework which is both compatible with its natural/autonomous computing model and rich to capture its complexity. The “complex system theory” is a meta-theory that provides a common modelling framework to study such complex systems under one umbrella. This meta-theory consists of a rich collection of tools adopted from various fields, where each tool is developed to study an instance of a complex system in a particular field, but applicable across other complex systems from different fields. In this paper, for the first time we introduce and apply the complex system theory as a modelling framework to PDNs. We demonstrate the usefulness of this modelling framework by presenting a case study, focused on the problem of efficient search in PDNs. Observing the similarity between PDNs and social networks, we adopt a model from the study of social networks to develop an efficient search mechanism for PDNs. More specifically, we propose the SWAM access method, a distributed index structure that enables efficient processing of various similarity search queries (namely, exact-match, range, and kNN queries) in indexable PDNs. SWAM is inspired by the “small-world” models originally introduced to explain efficient communication in social networks. We verify the efficiency of these search mechanism both analytically and empirically.