Research On Load Balancing Based On Peer-to-Peer Networks

Peer-to-Peer network(referred to as P2Pnetwork) is a combination product of adistributed system and computer network, and works based on peer mode. Comparedwith traditional networks, P2P networks have high scalability, high fault tolerance anddistribution, etc. Thus, they have wide application in sharing files, multimediatransmission areas, distributed data access and so on.These excellent features make it become the focus of the various universities andresearch institutions. With the increasement of network size, the surge in the number ofusers, load balancing problem in P2Phas become more prominent. The traditional loadbalance methods only considered single resource,butthere are multiple bottleneckresources in real network environment, such as storage capacity, ability of processingrequest and bandwidth, etc. Therefore, the research on multiple resources has a moretheoretical and practical meaning and faces to grater challenge. To address this issue, inthis paper, similarity concept model is established, then multiple resources load balance(MRLB) algorithm is developed for the arbitrary structured P2P network which usetransferred cost function defined for virtual server transfer. Simulation resultsdemonstrate that resource loads on a node can increase with the improvement of nodecapacity. So MRLB algorithm effectively solves the shortcomings of traditional methodsand achieves load balance with multiple bottleneck resources.Because of the excellent characteristics of P2Pnetworks, researchers combine theP2P network and streaming media services. The existing researches based on structuredP2P network can effectively reduce the delay, but not further guarantee the balance ofresources. We propose a new S-CQstructured P2Pnetwork topology model, whichcombines with the feature of crossed cubes, and establish the multiple independentspanning trees for data distribution in the layer. This network model has agood self-organization, delay, etc. The simulations also show that it can effectivelybalance the distribution of resources without increasing the average delay and the worstdelay, and improve the quality of playback.