An Adaptive Approach to Enhance the Performance of Content-Addressable Networks

Distributed hash table (DHT) algorithms structure peer-to-peer networks to provide nodes with fast and scalable lookups. In DHT solutions, such as Chord and Kademlia, the contacts of a node in the overlay network are determined so as to keep up with a lookup cost of O(logN) in a network of N nodes. As opposed to these, one of the first DHT solutions, called content addressable network (CAN), has the drawback of limiting the lookup cost only in O(dN^1/d) where d is the number of dimensions in the coordinate space, a fixed network parameter. However, CAN has several merits to exploit. Thus, in this paper we present a novel algorithm for CAN, called long-range CAN (LR-CAN) which is able to eliminate the rigidity of the original system, and provides a more scalable and resilient solution, not only compared to the original version, but to the currently best performing DHTs that we already mentioned.