Institutional Repository
| 网络拓扑和带宽实时度量技术研究 | |
| Alternative Title | Study of the Real-Time Measurement Approach on Network Topology and Bandwidth |
| 周辉 | |
| 2008-01-11 | |
| Degree Grantor | 中国科学院软件研究所 |
| Degree Level | 博士 |
| Place of Degree Grantor | 软件研究所 |
| Keyword | 网络性能 可用带宽 拓扑结构 主动探测 Icmp 全景快照 网络验证 |
| English Abstract | 在过去的二十年中，互联网取得了巨大的成功。互联网的成功，很大程度上归功于它简单易行的通信模式：一个节点只需要按照指定的协议发送和接收数据包，而不必了解负责传输数据的网络是如何运行的。互联网从设计之初至今，都没有显式地提供一种协议以支持节点去感知周围的网络。 尽管互联网的通信模式简单易行，但如果反过来要求一个节点去获知其所处网络的运行状态，那将是非常困难的。而最近五年中，越来越多的新需求开始涌现，迫切要求提高单一节点感知网络的能力。例如，网络管理员需要实时地度量局部网络的拥塞情况，从而升级或更新设备。安全工具需要能够快速感知网络中流量的异常变化，从而提供有效的预警。越来越多的流媒体传输应用，都要求感知端到端可用带宽的变化，从而实时调整数据传输策略，以期达到最佳的视频效果。我们的调查和实验都表明：传统的网络监控工具，已经无法满足网络管理和新型网络应用的需求。 开发单个节点对网络的感知能力，构造实时的网络性能度量系统，从而获知网络的状态参数，并优化对网络的使用方式，具有非常高的研究价值和应用前景。本文以主动探测技术为基础，从可用拓扑结构和带宽的度量入手，研究如何提高节点的网络感知能力，并在以下五个方面做出创新性贡献： (1) 度量端到端瓶颈属性的BNeck。提出了度量瓶颈链路的物理带宽、可用带宽及位置的算法，并据此实现了BNeck。这是第一个能够并发度量多个瓶颈属性的系统，我们的NS2 仿真和大规模网络实验都表明：BNeck 能够快速且准确地度量瓶颈链路的物理带宽、可用带宽和位置。 (2) 研究可用带宽度量的分类、原理及问题。在大量实验数据的基础上，分析了十三个常见度量系统的工作原理及软件实现，将其分为单双终端系统两类，进而总结了系统中普遍存在的十类基本问题。 (3) 拓扑结构度量工具TopDetect。为了获知网络的拓扑结构，我们开发了TopDetect。TopDetect 运行在单一节点上，通过发送IP 包触发路由器的ICMP 报错机制，从而获知周边存在的路由器，并据此构造网络拓扑。试验证明：在LAN 和WAN 中，TopDetect 都能实现高达80%的覆盖率。 (4) 网络拓扑和带宽的度量系统RichMap。RichMap 允许我们以图的方式浏览网络的拓扑结构，及其上所有链路的可用带宽。为了开发RichMap，我们整合了BNeck和TopDetect 中的数据探测功能，并优化探测数据包构建算法、ICMP 收集和绘图模块，从而确保拓扑探测和带宽度量高效地重叠在一起。 (5) 大规模网路试验及数据分析。分析在企业网和校园网内安装、调试和运行BNeck、TopDetect 及RichMap 时收集到的数据，得到很多有价值的结论。特别是分析了拓扑探测中采样偏差问题，反向路由中背景数据流对度量精度的影响，以及不同路由器缓存算法对探测数据包队列的干扰。 |
| Abstract | Characterizing the running status of the Internet has been proved to be a challenging task. Sincethe Internet is a collection of thousands of smaller networks, each under its own administrativecontrol, there is no single place from which one can obtain a complete picture of its topology. Moreover, because the design of network does not provide explicit support for direct inspection, the task of “obtaining” the surrounding network’s status has been left to researchers who developmore or less sophisticated methods to infer this topology from appropriate network measurements. Because of the elaborate nature of the network protocol suite, there are many possible measurements that can be made, each having its own strengths, weaknesses, and assumptions, and each leads to a distinct view of the same network. As a result, exploring the potential of node sensing ability, constructing a real-time network performance measurement system, and optimizing the ways we utilize networks, have been regarded as a promising research direction. Through a thorough study of topology discovery and bandwidth measurement, we discuss how to use active probing techniques to enhance the sensing ability of a single node. Specifically, we make the following five contributions. (1) The classification of 13 bandwidth measurement systems, as well as their pitfalls. (2) Developed BNeck, an end-to-end available bandwidth measurement system, which can measure the available bandwidth, capacity, and location of bottleneck. (3) Built TopDetect, a single-node tool which can automatically discover the topology, which was proved to be abile to cover up to 80% nodes and links of LAN and WAN. (4) Constructed RichMap, which outputs the network as a live map with measurement properties in a real-time manner. (5) A thorough analysis of the experimental data that we have been collected during our installation, debugging, and running BNeck, TopDetect, and RichMap. |
| Pages | 163 |
| Language | 中文 |
| Content Type | 学位论文 |
| URI | http://ir.iscas.ac.cn/handle/311060/5872 |
| Collection | 中科院软件所_中科院软件所 |
| Recommended Citation GB/T 7714 | 周辉. 网络拓扑和带宽实时度量技术研究[D]. 软件研究所. 中国科学院软件研究所,2008. |
| Files in This Item: | ||||||
| File Name/Size | DocType | Version | Access | License | ||
| 10001_20041801502901（2712KB） | 限制开放 | -- | Application Full Text | |||
Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.
Edit Comment