Research On The Credible And Collaborative Computing Model In P2P Environment

The development of the information age makes people explore the macro and micro unknown world into an unprecedented depth and breadth. The explosive growth of data size and the high complexity of the data scale have made the high-performance computing become a current hotspot of research. Due to the disadvantages of high costs, poor scalability and high energy consumption of traditional use of high performance computers to solve problem, people need a more affordable and efficient way to solve the problem of high-performance computing.Network technology, especially the rapid development of P2P technology, can make full use of most of the idle resources in the network. In addition, it maintains low costs and good scalability. It has become a good way to solve the problems we faced when traditionally using high performance computers to solve problems. Currently, these calculation projects are mainly divided into two categories:one is the desktop grid system, which accepts global volunteers from the network to join together to complete a scientific computing; the other is the cloud computing service, which integrates thousands of computers to form a super computer system to provide computing services for the outside world. Through P2P technology, both of the two categories above achieve a powerful computing capability. What’s more, they greatly enhance resource utilization, reduce energy consumption and provide effective services for high-performance computing in various fields.This paper analyzes the advantages of the existing ways of using P2P technology to solve high-performance computing problems and then continues to develop the existing advantages in resource utilization, scalability and operation. On this basis, it improves the existing P2P collaborative computing model from the aspects of improving the freedom and the credibility. This paper believes that nodes in the P2P network are equal with each other and the third party(including central servers and super nodes) is always incredible. Therefore, the collaborative computing model proposed in this paper is built in the fully distributed unstructured P2P network. In such environment, there is no third party and nodes can communicate with each other in quite a free way. The collaboration between nodes lies in:any node can hand out its computing tasks to other nodes and provide computing services for other nodes using its own resources in the leisure time.On the basis of improving the level of collaboration, it is important to form relatively complete trust model to provide protection for the results of computing services. This paper uses an improved D-S evidence combination theory and introduces recommend reputation into the trust model. When combining evidence, it gives different weights according to the recommended reputation. It filters out the evidence with large deviation and takes the valid information carried by the conflicting evidence into full consideration to compute the final trust value of the node.This paper uses PeerSim simulation platform to build experimental environment. It simulates the CCTM, EigenTrust and ETTM to compare the successful transaction rates of the three trust models under three malicious threats. The experimental results show that the CCTM maintains high successful transaction rats under malicious threats with good performance.