没有合适的资源?快使用搜索试试~ 我知道了~
温馨提示
试读
8页
通常,为了推断大规模网络的拓扑结构,研究人员使用有利位置进行广泛的基于跟踪路由的测量。 但是,推断的拓扑通常是不完整或不准确的,因为我们严重依赖的跟踪路由技术受到目标网络的规模和动态的限制。 研究采样偏差,即推断拓扑与实际拓扑之间的差异变得非常必要。 为此,我们捕获特定大型网络的近似完整拓扑,然后将完整拓扑与基于随机有利位置集的视图推断的拓扑进行比较。 我们发现,如果无法发现抽样偏差,则可能会严重破坏从推断拓扑得出的结论。 此外,采样偏差是特定于度量的。 它取决于估计的度量标准,并且与推断的拓扑可以覆盖的目标网络的百分比紧密相关。
资源推荐
资源详情
资源评论
Studying the Sampling Bias of Network Topology Discovery
Hui Zhou, WencaiDu
College of Information Science & Technology, Hainan University
Renmin Ave No. 58, Haikou, 570228, China
E-mail: h.zhou.china@gmail.com, [email protected]
Abstract
Usually, to infer the topology of a large-scale network, researchers use vantage points to conduct
extensive traceroute-based measurements. However, an inferred topology is often incomplete or
inaccurate because the traceroute technique we heavily rely on is limited by the scale and dynamics of
target network. It becomes very necessary to study the sampling bias, i.e. the difference between the
inferred topology and the real one. To do this, we capture an approximate complete topology of a
specific large-scale network, and then compare the complete topology with the inferred topologies that
are built on the views of a random set of vantage points. We find that sampling bias, if undetected,
could significantly undermine the conclusions drawn from inferred topologies. Furthermore, the
sampling bias is metric-specific; it depends on which metric is under estimation, and it is closely tied to
the percentage of target network that the inferred topology can cover.
Key Words: Network topology, sampling bias, topology discovery, topology metric
1. Introduction
Understanding the structural properties of the Internet has been proved to be a challenging
task. There is no single place from which one can obtain a complete picture of its topology
since the Internet is a collection of thousands of smaller networks, each under its own
administrative control. Moreover, because the design of network does not provide explicit
support for direct inspection, the task of “obtaining” the Internet’s topology has been left to
researchers who develop more or less sophisticated methods to infer this topology from a
large volume of network measurement data. Because of the elaborate nature of the network
protocol suite, there are many measurement approaches, each having its own strengths,
weaknesses, and assumptions, and each resulting in a distinct view of target topology.
In the last two decades, researchers have inferred five basic categories of network topologies.
They are the graphs of connections between autonomous systems (ASs) [2], the point-of-
presence (POP) topologies that interprets the structure of backbone using geography
information [15], the IP-level topologies whose nodes are IP addresses and whose links are
connections between the IP addresses [6, 7] the router-level topologies that resolve IP aliases
and group the IP addresses in the unit of router [19], and the connectivity of physical
资源评论
weixin_38644168
- 粉丝: 0
- 资源: 896
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功