Adaptive overlay construction through dynamic distributed matching with preferences

Rapport, 2011

An important function of overlay networks is the facilitation of connection, interaction and resource sharing between peers with limited resources. It is usually the case that the peers maintain some private notion of how a "desirable" peer should look like and they share their limited resources with peers that they prefer better than others. Recent research proposed that this problem can be modeled as a many-to-many matching problem with preferences and studied analytically. However, the proposed solutions studied the problem on static networks, where no node joining or leaving is considered. In this paper we present a dynamic, distributed algorithm for the many-to-many matching problem with preferences that utilizes a simple scheme, provides a guaranteed approximation for the total satisfaction in the network and guarantees the convergence after changes complete. We also provide a detailed experimental study of the algorithm that focuses on the levels of achieved satisfaction as well as convergence and reconvergence speed.