基于模糊综合评价的P2P网络流量优化方法研究

发布时间：2018-10-31 12:56

【摘要】：P2P (Peer-to-Peer Network)网络是构建在底层物理网络之上的逻辑覆盖网络,具有网络节点非中心化、高可扩展性、强健壮性、高性价比、负载均衡等多种特性。随着Internet的普及、网络接入带宽的增大以及端系统计算能力的迅速增强,P2P端到端通信模式已经逐渐发展成为IT行业和学术界应用和研究的热点。目前P2P技术在内容分发、流媒体、分布式存储、并行计算、云存储以及即时通讯等领域均有广泛应用,根据调查显示,国内主要运营商骨干网流量中P2P流量已达到了60%以上,成为因特网上的主要流量。然而多数P2P应用采取的随机邻居节点选取策略忽略了底层物理网络的信息,导致逻辑网络与物理网络严重失配,产生大量冗余跨域流量,加重了因特网服务提供商(ISP, Internet Service Provider)的流量压力并破坏了ISP收支平衡,同时P2P应用挤占带宽的局面对其他带宽敏感型应用也造成了不良影响。如何减少域间冗余流量,即对P2P流量进行优化,成为ISP以及P2P内容提供商共同关注的热点问题。目前一些P2P流量优化方法根据时延、带宽、AS距离等信息估计节点间网络距离,优先选取“邻近“节点作为邻居,然而这些方法大多选取单一因素进行网络距离度量,在复杂的因特网背景下这种方式无法保证有效性。本文提出了一种基于模糊综合评价的P2P流量优化方法,当某个P2P节点进行邻居节点选择时综合考虑时延、带宽、AS距离、链路代价、文件差异程度等多种因素对所有节点进行评价并排序,从中选取评价结果较高的P2P节点作为该节点的邻居节点,据此构建逻辑覆盖网络。该方法采用结合粒子群算法(PSO, Particle Swarm Optimization)的改进层次分析法(AHP, Analytic Hierarchy Process)确定各因素权重,同时验证和修正一致性判断矩阵。本文还对提出的优化方法进行了网络模拟实验,分析结果表明该算法能够大幅度本地化P2P流量并且一定程度上加速文件下载,保证P2P应用性能不下降的同时进一步优化P2P流量。
[Abstract]:P2P (Peer-to-Peer Network) network is a logical overlay network built on the underlying physical network. It has many characteristics, such as non-centralization of network nodes, high scalability, strong robustness, high performance-to-price ratio, load balancing and so on. With the popularity of Internet, the increase of network access bandwidth and the rapid enhancement of end-to-end computing power, P2P end-to-end communication mode has gradually developed into a hot spot in the application and research of IT industry and academia. At present, P2P technology is widely used in the fields of content distribution, streaming media, distributed storage, parallel computing, cloud storage and instant messaging. According to the survey, P2P traffic in the backbone network of major domestic operators has reached more than 60%. Become the main traffic on the Internet. However, the random neighbor node selection strategy adopted by most P2P applications ignores the information of the underlying physical network, resulting in a serious mismatch between the logical network and the physical network, resulting in a large number of redundant cross-domain traffic, and exacerbating the Internet service provider (ISP,). Internet Service Provider) traffic pressure and break the balance of ISP revenue and expenditure, while P2P applications of bandwidth crowding out of other bandwidth sensitive applications also have a negative impact on other applications. How to reduce redundant inter-domain traffic, that is, to optimize P2P traffic, has become a hot issue for ISP and P2P content providers. At present, some P2P traffic optimization methods estimate the network distance between nodes according to the information of delay, bandwidth, AS distance, and select "adjacent" nodes as neighbors first. However, most of these methods choose a single factor to measure the network distance. This approach does not guarantee effectiveness in a complex Internet context. In this paper, a P2P traffic optimization method based on fuzzy comprehensive evaluation is proposed. When a P2P node selects a neighbor node, time delay, bandwidth, AS distance, link cost are taken into account. All nodes are evaluated and sorted by many factors such as file difference degree and so on. P2P nodes with higher evaluation results are selected as neighbor nodes of this node and the logical overlay network is constructed according to these factors. In this method, the weight of each factor is determined by the improved Analytic hierarchy process (AHP, Analytic Hierarchy Process) combined with particle swarm optimization (PSO, Particle Swarm Optimization), and the consistency judgment matrix is verified and amended. The results show that the proposed algorithm can localize P2P traffic greatly and accelerate file download to a certain extent so as to ensure that P2P application performance does not deteriorate and optimize P2P traffic further.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.02;TP393.06

【参考文献】