基于随机规划的IP流量矩阵估计方法的研究

发布时间：2018-05-20 09:25

本文选题：流量矩阵 + 层析成像　；参考：《华中师范大学》2015年硕士论文

【摘要】：互联网技术是现如今发展速度最快、应用最广泛的技术之一,然而在近些年来,随着互联网技术逐渐的成熟、互联网应用的进一步普及,大量的新型网络服务和应用在互联网中如雨后春笋般涌现出来。巨大的网络规模、大量的链路传输数据、许多异构网络的接入,使得网络研究人员对网络直接进行网络测量来获得流量矩阵已经非常困难。流量矩阵是许多网络技术的重要支撑,它反映的是每个网络路径中的流量需求,在许多工程领域中有着重要的应用。我们必须寻求一种新的方式来解决面临的问题,能够有效、快速获取到网络中的流量矩阵,为下一步的网络研究打好基础。在本文中详细介绍了流量矩阵概念以及其获取方法。对流量矩阵估计问题的方法及相关模型进行了总结和概括,重点介绍了层析成像技术和重力模型。流量矩阵用直接测量的方式是行不通的,只有通过估计的方式去获取,本文的重点任务就是要解决克服流量矩阵估计问题。在流量矩阵估计方程中,OD流的数目远大于IP网络中的链路数,所以导致这个方程为欠定的、病态的方程,求解起来非常困难。并且以往所有的模型是在理想的情况下进行的,没有考虑链路噪声的存在。本文为此就提出了一个新的模型—随机规划模型(Stochastic Programming Model)。通过在流量矩阵估计方程的约束函数中引入随机变量,使原来的等式方程变为概率求解方程,增大了方程的求解空间,从而增大寻求最优解的可能性,并且最关键的是,用随机变量代表网络中的链路噪声,能够更好地模拟现实的网络环境。通过全面的理论分析和基于真实网络数据,并与经典的层析成像重力模型(Tomogravity Model)作对比的仿真实验,结果我们可以看出,随机规划模型的估计效果更好,与网络的实际值更加接近。
[Abstract]:Internet technology is one of the fastest growing and most widely used technologies nowadays. However, in recent years, with the maturity of Internet technology, Internet applications have become more and more popular. A large number of new network services and applications have sprung up in the Internet. Because of the huge network scale, the large amount of link transmission data, and the access of many heterogeneous networks, it is very difficult for network researchers to measure the network directly to obtain the traffic matrix. Traffic matrix is an important support of many network technologies. It reflects the traffic requirements in each network path and has important applications in many engineering fields. We must find a new way to solve the problem, which can effectively and quickly obtain the network traffic matrix, and lay a good foundation for the next network research. In this paper, the concept of flow matrix and its acquisition method are introduced in detail. The methods and models of flow matrix estimation are summarized, and the tomography technique and gravity model are introduced. It is not feasible to measure the flow matrix by direct measurement. The main task of this paper is to overcome the problem of estimating the flow matrix. The number of OD flows in the flow matrix estimation equation is much larger than the number of links in the IP network, so it is very difficult to solve this equation because it is an ill-defined and ill-defined equation. And all previous models are carried out under ideal conditions without considering the existence of link noise. In this paper, a new model, stochastic Programming model, is proposed. By introducing random variables into the constraint function of the flow matrix estimation equation, the original equation is transformed into a probabilistic solution equation, which increases the space for solving the equation and increases the possibility of seeking the optimal solution. Using random variables to represent the link noise in the network can better simulate the real network environment. Through comprehensive theoretical analysis and simulation experiments based on real network data, and compared with the classical tomogravity model, we can see that the stochastic programming model is more effective than the classical tomogravity model. Closer to the actual value of the network.
【学位授予单位】：华中师范大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP393.06

【共引文献】