P2P环境下网络流量管控技术研究

发布时间：2018-05-07 08:49

本文选题：对等网络 + 深度数据包检测　；参考：《南京邮电大学》2014年硕士论文

【摘要】：对等网络突破了传统C/S或B/S等不对称计算模式，每个结点地位对等，，可同时成为服务的使用者和提供者，从而为大规模的信息共享、直接通信和协同工作提供了灵活的可扩展计算平台。但随着P2P用户量的增多，主干网络中P2P流量的比例也逐渐变大，严重影响其他网络业务的运行，这使得P2P应用技术的无限需求与有限主干网络带宽之间产生了矛盾。为解决这一矛盾，较有效的方法是对P2P的流量进行合理的疏导，在主干网边缘架设P2P缓存系统，以将P2P流量尽量控制在内网之中，从而缓解主干网的压力。本文从P2P流量管控的角度进行研究，首先介绍了P2P技术以及网络拓扑结构，并对几种常用的流量识别方法以及典型的P2P协议的通信过程和关键数据包的格式进行分析，提出了基于自主学习的P2P流量识别方法，此方法充分利用了DPI、DFI以及神经网络算法的各自特点，提高了P2P流量的识别准确率。然后，在上述识别方法的基础上，通过结合NetFilter架构、Socket Buffer、NetLink等技术，设计了P2P流量管控系统，并对系统的五大模块：流量识别模块、数据交互模块、报文缓存模块、数据包伪造模块、流量统计模块分别进行了详细的设计和实现。最后，论文对所实现的P2P管控系统进行了说明，介绍了相关的软硬件平台以及安装编译过程，整个P2P流量管控系统达到了管理和控制P2P流量的效果。
[Abstract]:Peer-to-peer networks break through the traditional asymmetrical computing models such as C / S or B / S, each node is equal in status, it can become the consumer and provider of service at the same time, thus sharing information on a large scale. Direct communication and collaborative work provide a flexible scalable computing platform. However, with the increase of the number of P2P users, the proportion of P2P traffic in the backbone network is gradually increasing, which seriously affects the operation of other network services, which makes the unlimited demand of P2P application technology and the limited backbone network bandwidth conflict. In order to solve this contradiction, the more effective method is to reasonably channel the P2P traffic and set up a P2P cache system on the edge of the backbone network, so as to control the P2P traffic in the inner network as far as possible so as to relieve the pressure on the backbone network. From the point of view of P2P traffic management and control, this paper first introduces P2P technology and network topology, and analyzes several commonly used traffic identification methods as well as the communication process of typical P2P protocols and the format of key data packets. A P2P traffic identification method based on autonomous learning is proposed, which makes full use of the characteristics of DPI DFI and neural network algorithms, and improves the accuracy of P2P traffic recognition. Then, on the basis of the above identification methods, a P2P traffic control system is designed by combining with the technology of socket buffer NetLink and NetFilter architecture. The five modules of the system are: traffic identification module, data exchange module, message buffer module, etc. Packet forgery module and traffic statistics module are designed and implemented in detail. Finally, the paper describes the P2P management and control system, introduces the related software and hardware platforms and the installation and compilation process, the whole P2P traffic control system has achieved the effect of managing and controlling P2P traffic.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.02

【参考文献】