Beyond Node Degree: Evaluating AS Topology Models

arXiv:0807.2023v1 [cs.NI] 13 Jul 2008 Beyond Node Degree: Evaluating AS Topology Models Hamed Haddadi ∗ University College London Damien Fay University of Cambridge Almerima Jamakovic Delft University of Technology Olaf Maennel † Deutsche Telekom Laboratories Andrew W. Moore University of Cambridge Richard Mortier ‡ Vipadia Ltd Miguel Rio University College London Steve Uhlig Delft University of Technology ABSTRACT Many models have been proposed to generate Internet Au- tonomous System (AS) topologies, most of which make struc- tural assumptions about the AS graph. In this paper we com- pare AS topology generation models with several observed AS topologies. In contrast to most previous works, we avoid making assumptions about which topological properties are important to characterize the AS topology. Our analysis shows that, although matching degree-based properties, the existing AS topology generation models fail to capture the complexity of the local interconnection structure between ASs. Furthermore, we use BGP data from multiple vantage points to show that additional measurement locations signif- icantly affect local structure properties, such as clustering and node centrality. Degree-based properties, however, are not notably affected by additional measurements locations. These observations are particularly valid in the core. The shortcomings of AS topology generation models stems from an underestimation of the complexity of the connectivity in the core caused by inappropriate use of BGP data. Categories and Subject Descriptors C.2.1 [Network Architecture and Design]: Net- work topology; I.6.4 [Simulation and Modeling]: Model Validation and Analysis General Terms Topology, Models, Measurement ∗This work was done while the author was visiting the Com- puter Laboratory, University of Cambridge. †This work was done while the author was at School of Mathematical Sciences, University of Adelaide. ‡This work was done while the author was at Microsoft Re- search Cambridge. Keywords Internet, BGP, Topology generation, Graph metrics 1. INTRODUCTION For many years researchers have modeled the Inter- net’s Autonomous System (AS) topology1 using graphs obtained via various measurement techniques, e.g. BGP routing tables [16, 28] and traceroute maps [18]. The AS topology is an abstraction of the Internet which is commonly used to analyze its characteristics and sim- ulate the performance and scalability of new protocols and applications. Simulation methods require that AS topology generation models be able to provide topolo- gies whose properties are as close as possible to those of the observed AS topology. In this paper we evaluate existing AS topology gener- ation models by comparing them with several available datasets, representing observed AS topologies of the In- ternet. Figure 1 illustrates the relationship between the Internet topology, its measurement instances, and AS topology generation models. A key principle underlying our work is to be agnostic about the topological properties of the Internet. The main reason for our agnosticism lies in the dynamic be- havior of the Internet topology. In addition, observa- tions of the AS topology suffer from two problems. On the one hand, common set of observation points have only limited visibility of the topology [26]. On the other hand, each observation technique suffers from measure- 1Note that the AS topology neither represents the data- plane topology nor directly corresponds to the Internet router-level topology. Many organizations are permanently connected to their providers, sharing an AS number [29]. Al- ternately, a single organization may use many AS numbers for controlling routing. 1 Figure 1: Internet topology generation ment artifacts. This results in problems for BGP-based as well as traceroute-based observations of the Internet topology. For example, traceroute can report hops that do not map to a unique AS number [22]. As a result, AS topology models make use of simplifying assump- tions about the actual topology [5, 19, 37]. One widely held assumption, based on biased observations, is that the AS topology has a hierarchical structure [30] and its node-degree distribution obey a power-law [12]. Believing that at present it is impossible to know bet- ter, we accept the fact that the AS topology observa- tions suffer from biases and thus reveal different partial truths about the properties of the Internet. However, comparison of different observed AS topologies with dif- ferent levels of incompleteness, and topologies generated from different models, allows us to learn from the lim- itations of particular assumptions about the Internet’s AS topology. Then, the direction of these biases and limitations may gives us insight into the actual proper- ties of the AS topology. To evaluate AS topology generation models, we rely on a wide set of commonly used topological metrics. We do not claim that the set of considered metrics captures all important aspects of

…(Full text truncated)…

This content is AI-processed based on ArXiv data.