Dissertation > Excellent graduate degree dissertation topics show

Research on Distributed Cache Strategies in Peer-to-Peer Networks

Author: GaoGuoQiang
Tutor: LuZhengDing；LiRuiXuan
School: Huazhong University of Science and Technology
Course: Applied Computer Technology
Keywords: Peer network Topology Probability Forwarding Distributed Cache Proactive Caching Search for scarce resources
CLC: TP393.02
Type: PhD thesis
Year: 2011
Downloads: 192
Quote: 1
Read: Download Dissertation

Abstract

Peer network (Peer-to-Peer Network, P2P) is a distributed network, built on top of the physical network logical overlay network, each computer on the network has the same function, mutual sharing of resources. Although the large number of studies, but due to the complexity of network structure, and other network resources on a relatively poor search efficiency problem still exists. Meanwhile, with the rapid growth of P2P applications, peer to peer network system has become the Internet's largest bandwidth consumers. Improve P2P search efficiency, reduce Internet backbone traffic pressure, increase the user's access experience, P2P applications have become an important issue to be solved. Caching technology to solve these problems is an effective approach. Although based on a central server, centralized caching strategies have been developed, but for distributed caching mechanism P2P systems research is still in its infancy. The basic unstructured P2P network search mechanism is flooding, such a strategy has a low search efficiency. To improve search efficiency and reduce system overhead, we propose a distributed unstructured P2P network based caching mechanism, including location-based topology structure, three-dimensional random walk preference, maximum delay cache placement policy, based on heat and requests cache replacement policy. In the topological structure, the aggregate position similar nodes to form different subnets; search is biased in favor of the subnet search query to get a shorter delay; caches placed in the decision node, select the node farther away from the target nodes to improve the cache hit rate; replaced in the cache, replacing high priority resources, prevent excessive cache problem. Through simulation experiments on the proposed algorithm is simulated and compared with other algorithms analysis, the results show that the proposed algorithm has better query hit rate, the smaller the query latency, higher cache hit rate and lower communication overhead. Structured P2P networks in order to improve the overall performance, we propose a structured P2P network based distributed caching method. By using the hierarchical distributed hash tables (Distributed Hash Table, DHT) routing structure, can have a different rate (the resource in the network and the number of copies of the network size ratio) of the various resources to build a sufficient number of cache, and carefully select the appropriate node to store the cache to improve cache efficiency. Through the different algorithms and heuristic algorithm simulation experiment found that, combined with hierarchical DHT routing structure, the proposed caching strategy can effectively reduce query latency for better load balancing and better cache hit rate, but also effectively alleviate the problems caused by excessive negative caching and the impact, most of the resources can be better to provide satisfactory cache performance. Most P2P distributed caching mechanisms are passive caching, that is, only when a resource is requested to trigger cache operation, which will result in insufficient cache scarce resources, scarce resources are being requested because the frequency is relatively low. However, in many cases, is very valuable scarce resources, such as P2P file sharing and P2P streaming media applications, a rare piece. In order to improve the search efficiency of scarce resources, proactive caching strategies proposed search algorithm and scarce resources, it requires less number of cached copies, which can reduce the overhead caused by caching mechanism. Simulation results show that by the proposed algorithm can greatly improve the search efficiency of scarce resources, the search hit rate of 98% or more; the communication overhead while maintaining a lower level compared to the existing algorithms can be reduced Nearly 40% of the traffic. P2P-based streaming and file sharing has become increasingly popular, and the scale of these applications is also growing, increasingly important impact on the Internet. However, due to the existing large-scale commercial systems are mostly P2P applications, these applications protocols, architectures and algorithms are mostly not public, making analysis of these systems, identify problems and resolve bottlenecks becomes very difficult. Numerous studies using the black box approach to measure and assess the characteristics of these systems, such as download performance, user distribution, session length. In order to cover more measurements, the study measured peer network, and presents a generic peer network measurement model and the use of resources for different heat proposed model was validated experiments. Measurement scale including data transfer rate, the cache mechanism, node locations distributed, session length, these measurements can be distributed cache to support research, but also for other researchers to help.

Related Dissertations

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Computer network > General issues > Computer network architecture and design
© 2012 www.DissertationTopic.Net Mobile