Improving the scalability of logarithmic-degree DHT-based peer-to-peer networks

High scalability in Peer-to-Peer (P2P) systems has been achieved with the emergence of the networks based on Distributed Hash Table (DHT). Most of the DHTs can be regarded as exponential networks. Their network size evolves exponentially while the minimal distance between two nodes as well as the routing table size, i.e., the degree, at each node evolve linearly or remain constant. In this paper we present a model to better characterize most of the current logarithmic-degree DHTs. We express them in terms of absolute and relative exponential structured networks. In relative exponential networks, such as Chord, where all nodes are reachable in at most H hops, the number of paths of length inferior or equal to H between two nodes grows exponentially with the network size. We propose the Tango approach to reduce this redundancy and to improve other properties such as reducing the lookup path length. We analyze Tango and show that it is more scalable than the current logarithmic-degree DHTs. Given its scalability and structuring flexibility, we chose Tango to be the algorithm underlying our P2P middleware