The law of the distribution in KAD's routing table

Abstract

In recent years, in the field of file sharing on the Internet, BitTorrent and eMule accounted for a large proportion of the application. The two software have their own proprietary peer-to-peer (P2P) network, and are based on the Kademlia protocol development, Kademlia protocol is typical of the DHT(Distributed Hash Table) protocol. In eMule, proprietary peer-to-peer (P2P) network is called the KAD network, the goal to detect KAD network is to discover the relationship between the nodes and find new nodes in network rapidly and completely. Optimization of detection, we think, should be based on the distribution in routing table. Our job is to detect some nodes in KAD, and find out the existing laws of routing table. At present, only a few articles that involves detection of the routing table, but there is not enough analysis to summarize laws. In our experiment, We have developed a crawler, called V-CRAWLER. We used it to crawl the KAD, and found one million and five hundred thousand nodes in one time. We have detected some of their routing table, found out that their routing table is not always in full. The detection results show that the actual routing table structure of KAD network nodes has clustering property. The neighboring nodes center on the upper 20 layers of the similar binary tree structure of the routing table, and the first 15 layers occupy more than 85% of the nodes.