Peer-to-peerSystem.PDFVIP

LinkNet: A New Approach for Searching in a Large Peer-to-peer System Kunlong Zhang, Shan Wang School of Information, Renmin University of China, Beijing, China, 100872 zhangkl@, suang@ Abstract. Searching a file by its name is an essential problem of a large peer- to-peer file-sharing system. Napster and Gnutella have poor scalability. DHT- based systems do not keep the order of the keys. Skip graphs and SkipNet have too many links. In this paper, we present a new scalable distributed data struc- ture LinkNet for searching in a large peer-to-peer system. In LinkNet, all ele- ments are stored in a sorted doubly linked list, and one node stores many ele- ments. LinkNet uses virtual link to speed search and enhance fault tolerance. Because LinkNet is based on a sorted list, it benefits operations such as range query, bulk loading of data, and merging of two LinkNets. 1 Introduction For a large peer-to-peer file-sharing system, searching a file by its name is an essen- tial problem. One initial approach is to setup up a server which maps a file name to its location. Napster ([6]) uses this approach. The problem of this approach is that it uses an unscalable central database to index all files. Another initial approach is that a node broadcasts the search request to all its neighbors when it does not find the file in its local database. Gnutella ([7]) adopts this approach. However this approach doesnt scale well because of its bandwidth consumption and unrelated search in many nodes. To overcome the scalability problem, several algorithms based on a distributed hash table (DHT) approach are presented ([1], [8], [9], [10], [11]). In these algorithms, each node in the system maintains a small routing table to