Research Of DHT Distributed Search Algorithm Based On Chord

Because peer-to-peer networks have advantage in decentralization, scalability, robustness, cost-effective, load balancing, privacy protection and special architecture, it grows rapidly recently. Structured networks using DHT (distributed hash table) technology to locate the storage of nodes in a distributed system. The main applications of the distributed hash table are CAN, Tapestry, Pastry, Chord. This paper focuses on the Chord. Chord is a structured method to retrieve information through the distributed hash table (DHT) technology. DHT use consistent hash to put the query keywords and node identifier (id) in the same address space. At the same time, Chord holds a Routing table called finger table, so that each node only need to maintain a small number of nodes while keeping efficient lookup. Chord routing algorithm has been proved load balancing, robustness, scalability, availability, flexibility, but it also has drawbacks caused by the heterogeneity of nodes, low efficiency of nodes joining and leaving, difficulties on implementing fuzzy queries technologies.The way to improve Chord algorithm mostly by means of modifying the structure of finger table or improving the method of lookup. At present, typical improved algorithm of the Chord protocol are the F-Chord algorithm, PNS algorithm (the Proximity neighbor selection), and parallelism algorithm. The improved algorithm which uses a modified structure of finger table decreased the average hops and latency by increasing the size of fingertable, which leads Chord to cost more bandwidth in maintaining。This paper analyzed the lookup properties of Chord and proposed an improved algorithm CH-Chord, which based on information replication and hot-spots in looking up.This paper used a simulator named P2Psim to prove that CH-Chord can decrease the hops and latency without additional bandwidth cost by simulation. In additional, as a result of using parallel query, CH-Chord decreased the number of failed lookups because it can decrease the probability of congested network between lookup nodes and the storage nodes.