Measurement And Application Of Kademlia DHT Highly Available Peers

The Kademlia DHT has been an active research area since Kademlia protocol was proposed in2002. Various types of systems and application software who supports Kademlia protocol have been developed. The Kademlia DHT network is always consisting of millions of simultaneous users. This makes Kademlia become one of the most widely used file-sharing network protocols. Just like other P2P systems, the dynamic participation of peers in Kademlia DHT affects the performance of the networks. Fortunately, some peers in Kademlia DHT are always online and available. They are actively participating in P2P and selflessly sharing their resources with others. They are the core of the whole network. We call them highly available peers. The research of this thesis will be launched based on the features of these highly available peers, directly against several core issues in the current study as follows:· Peer Online Behavior Analysis and Prediction:In the DHT network, the participation of the network nodes directly determines the performance of the entire network. Peer Online Behavior Analysis and Prediction has been the focus in the research area.· Promotion of Content Publishing and Retrieval:Kademlia is one of the most widely used file-sharing network protocol. How to use measurement results to improve content publish-ing and retrieval performance has been a concern in the measurement study of Kademlia DHT network. Some lookup in content search fails after content has been published, al-though the existing network supports millions of users for content publishing and retrieval. The existing performance can be improved.· Service Based on Kademlia DHT Network:in Kademlia DHT network, participating peers contribute their own bandwidth, storage, computing and other network resources together to form the current file-sharing networks. How to make a better use of these network resources is a new research direction.To solve the above problems, this thesis carried out measurements on KAD and Mainline DHT. The content of this thesis includes three issues:(1) how to get highly available peers in Kademlia DHT;(2)improvement on content publishing and retrieval based on highly available peers;(3) web cache system based on stable peers in Kademlia DHT. The works in this thesis has been supported by National Science Foundation of China No.61301127; No.61371080and No.61370231. And it also has been supported by the Technology Support Plan of the National ’Eleventh Five-Year-Plan’of China under Grant No.2011BAK08B01and No.2011BAK08B02-01.The contributions of this thesis are as follows: 1. A Mechanism to Get Highly Available Peers in Kademlia DHT:From the measurement results of KAD and Mainline DHT, we find that the probability of peer ID appearance in routing tables can be used to characterize peers’availability. We give an algorithm of locating highly available peers in Kademlia DHT. The peers obtained by this algorithm is more stable, has a high online probability.2. A Content Publishing and Retrieval Algorithm Based on Highly Available Peers:This thesis measures and analyses the process of content publishing and retrieval in Kademlia DHT. An algorithm based on highly available peers is proposed to improve the perfor-mance of content publishing and retrieval.3. Web Cache System based on Stable Peers in Kademlia DHT:The dynamic participation of peers affects the performance of web cache system based on Kademlia DHT. Based on highly available peers, a stable peers’ selection algorithm is proposed to select stable peers in Kademlia DHT as web cache peers to improve the system performance.