Achieving and Maintaining Cost-Optimal Operation of a Hierarchical DHT System

Although hierarchical P2P systems have been found to outperform flat systems in many respects, current P2P research does not focus on strategies to build and maintain such systems. Available solutions assume either no or little coordination between peers, that could lead the system toward satisfying a globally defined goal (e.g., minimizing traffic). In this paper we focus on hierarchical DHTs and provide a full set of algorithms to build and maintain such systems, that mitigate this problem. In particular, given the goal state of minimizing the total traffic without overloading any peer, our algorithms dynamically adjust the system state as to keep the goal met at any time. The algorithms are fully decentralized and probabilistic, all decisions taken by the peers are based on their partial view on a set of system wide parameters. Thus, they demonstrate the main principle of self-organization - the system behavior emerges from local interactions. Our simulations, run in a range of realistic settings, confirm a good performance of the algorithms.