Research And Improvement Of P2P Routing Algorithm Based On Chord

In recent years, Peer-to-Peer quickly became one of the hot topics in the computer science. Fortune magazine even listed the P2P as one of the four technologies which will change the future of Internet.The core mechanism of P2P network is to establish a logical overlay network above the application layer. Since2001, academics put forward the most representative models of structured P2P network, such as Chord、CAN、Tapestry and Pastry. Structured P2P network builds a strict topology of the overlay network over application layer, and usually uses the distributed hash table (DHT) based on the consistent hash function. The DHT efficiently maps the network nodes or data objects onto the overlay network.The paper also realized that the topology of overlay network was not always consistent with the topology of the underlying physical network, so the routing didn’t result in a good physical delay.In order to avoid high-latency hop, we improved the actual overlay network routing performance by integrating the topology information of the underlying physical network with the overlay network. The paper analyzed the traditional Chord algorithm, found that the original Chord routing table based only on the neighboring information of the node ID over the overlay network, and the choice of the Chord routing table entry was very flexible, the ith Chord finger table entry of the node with ID n properly referred to the ID-space range n+2i-1to n+2i-1, the total choices were2i-1. In this paper, we use the heuristic PNS(K) method, the PNS(K) algorithm considers up to the first K nodes in that range (there may be fewer than K), and the ith finger table entry points to the lowest-latency node in the entire allowed ID-space range.Then the problem is how to collect information of the physical network topology. After analyzing the methods of integrating the underlying physical network topology information, the paper selected the distributed binning scheme proposed by Ratnasamy et al., whereby nodes partition themselves into bins such that nodes that fall within a given bin are relatively close to one another in terms of network latency. The scheme requires a set of well-known landmark machines spread across the Internet. An application node measures its distance, i.e. round-trip time, to this set of well known landmarks and independently selects a particular bin based on these measurements.The p2psim simulator developed by MIT is used to obtain the quantitative analysis data. Because we integrate the underlying physical topology information into the routing table and each hop, the routing performance improves.