A Network Topology Based Resource Discovery Model And Search Technology

Traditional resource sharing uses C/S mode, the server performance restricted the overall performance of the network, and a lot of client resources are not fully utilized, resulting in a waste of resources. How to take full advantage of all the idle resources,and avoid server bottlenecks is the focus of research about resources Sharing and resource discovery.Centralized information Index model records resource information to do resource discovery, This method makes use of the resources at "margins" of network, But there are also server bottlenecks and single points of failure.Gnutella established a wide distribution resource sharing network, Flooding to the neighbors to sent resources inquiries information, information can quickly spread to a number of nodes, but the network bandwidth consumption soon in Flooding way, and is inefficient to discover resources. Distributed Hash Table(DHT) adopted Distributed hash function, the only keys mapped to a particular node.Then use some routing algorithm to establish links connecting the nodes. This method is more efficient, but the way DHT maintain mechanism is rather complicated, exists the problem of excessively high maintenance costs.After analysis of the realistic network, it is found that many networks such as Gnutella network topology and social network both have a small world and the power-law properties. A small world network has a high degree of aggregation characteristic and exist a shorter path between two nodes. Power-law characteristic statement of a few network nodes have higher degree, and more links to other nodes.Most node degrees lower.Globus, Gnutella and other resources discovery modes de not consider these network topology characteristics. Take advantage of these characteristics, Network can be divided into different "clusters" according to the nodes interests.nodes have similar interest are organized together to create interest clusters, the resource discovery will search in the cluster first. Because of the similar interest in cluster, the success rate of resource discovery will greatly increase. When clusters are not satisfied, then request will be transmitted to a higher degree and more similar node according to Power-Law property.