Proximity neighbor selection using IP prefix matching in Kademlia-based Distributed Hash Table

KAD is one of the most popular Peer-to-Peer (P2P) networks on the Internet. It uses Kademlia-based Distributed Hash Tables (DHTs) to form a tree-structured P2P network. Iterative routing algorithm is used on Kademlia to perform key-value DHT lookup operation based on XOR distance of peer identifiers, which are randomized. Therefore the lookup operation might increase the lookup latency, as well as introduce a lot of cross-network traffic to other Internet Service Providers (ISPs). In this paper, we propose a new efficient proximity neighbor selection in Kademlia-based DHT. The locality among the peers is simply justified based only on the IP prefix matching of peers´ IP addresses. Unlike other existing approaches, there are no needs of local database of geographic information about peers, or modification of peer identification structure. Results from simulation on 2000 nodes network show that the proposed technique can reduce both the lookup latency and cross network traffic significantly, comparing to the traditional Kademlia lookup algorithm.