Research Of Data Location And Replication Algorithms In P2P-based Distributed Storage System

Network has become more and more important in modern society with its rapid development and popularization, and is still getting larger and larger. As a result, more and more storage resources of different types and quantities are joining the network, but these great amounts of distributed storage resources are not fully utilized. Fortunately, some new network technoligies have made it possible to make efficient use of them. Nowadays, it is a widely concerned problem in how to combine the distributed heterogeneous storage resources into one large distributed storage system which can provide secure and high-quality storage services.P2P(Peer-to-Peer) can reasonably and fully utilize all sorts of network resources. In a P2P-based system, there is neither server nor client, all nodes are symmetric and can communicate directly with each other. The distribution, equality, autonomy, scalability, flexibility and dynamic adaptation of P2P mode can do favor of the management of distributed storage system. By way of fully distributed management, P2P-based distributed storage system aims to make better use of storage resources and construct high-performance storage systems with each part in it being equal to each other.This paper concentrates on data location and replication algorithms in P2P-based distributed storage system which can be categorized into two types, centralized system and decentralized system which can further be divided into unstructured decentralized system and structured decentralized system. In the area of unstructured decentralized system, firstly some existing location algorithms and related technologies are analyzed, then the hop location algorithm is designed to collect information of nearby nodes and use this information to shorten user response time, and reduce the amount of network bandwidth and node resources used. In the area of structured decentralized system, at the beginning, some famous structured decentralized systems are analyzed, and then, by considering the factors that can influence the performance of the location algorithm, the fission location algorithm is adopted and realized. In fission, nodes are connected into one Kautz-liked topology. The degree of each node is constant, and the load of each node is close to each other. Fission system is easy to manage, and can offer load balance and dynamic adaptation. What's more, it is easy to build applications on top of it. Based on fission algorithm, the near replication algorithm is presented, which can achieve low user response time and load balance. At last, we simulate all the algorithms mentioned above to test their properties, performance and validity.