## A quantitative comparison of graph-based models for internet topology (1997)

### Cached

### Download Links

- [www.ssfnet.org]
- [www.cc.gatech.edu]
- [cs.baylor.edu]
- DBLP

### Other Repositories/Bibliography

Venue: | IEEE/ACM TRANSACTIONS ON NETWORKING |

Citations: | 235 - 3 self |

### BibTeX

@ARTICLE{Zegura97aquantitative,

author = {Ellen W. Zegura and Kenneth L. Calvert and Michael J. Donahoo},

title = {A quantitative comparison of graph-based models for internet topology},

journal = {IEEE/ACM TRANSACTIONS ON NETWORKING},

year = {1997},

volume = {5},

pages = {770--783}

}

### Years of Citing Articles

### OpenURL

### Abstract

Graphs are commonly used to model the topological structure of internetworks, to study problems ranging from routing to resource reservation. A variety of graphs are found in the literature, including fixed topologies such as rings or stars, "well-known" topologies such as the ARPAnet, and randomly generated topologies. While many researchers rely upon graphs for analytic and simulation studies, there has been little analysis of the implications of using a particular model, or how the graph generation method may a ect the results of such studies. Further, the selection of one generation method over another is often arbitrary, since the differences and similarities between methods are not well understood. This paper considers the problem of generating and selecting graph models that reflect the properties of real internetworks. We review generation methods in common use, and also propose several new methods. We consider a set of metrics that characterize the graphs produced by a method, and we quantify similarities and differences amongst several generation methods with respect to these metrics. We also consider the effect of the graph model in the context of a speciffic problem, namely multicast routing.

### Citations

1962 | On the self-similar nature of Ethernet traffic - Leland, Taqqu, et al. - 1994 |

1934 | Random Graphs
- Bollobás
- 1985
(Show Context)
Citation Context ...ter r de nes the boundary: P (u; v) = if d<r if d r One nice feature of the Locality method is that we have been able to extend some analytic results from the large body of work on pure random graphs =-=[10]-=- to this model; these results deal with properties of the graphs (e.g. connectedness) as the number of nodes becomes large. To summarize, Table 1 indicates the probability ofan edge between two vertic... |

1574 | Wide-area Traffic: The Failure of Poisson Modeling - Paxson, Floyd - 1995 |

1513 | Reducibility among combinatorial problems - Karp - 1972 |

1370 |
Graph Theory with Applications
- Bondy, Murty
- 1976
(Show Context)
Citation Context ...n nodes and m edges, we consider the following topological properties: Average node degree. (2m=n). Diameter. The diameter of a graph is the length of the longest shortest-path between anytwovertices =-=[11]-=-. Informally, alow diameter generally corresponds to shorter paths. We consider several variations on the diameter, depending on the metric used to construct and evaluate the shortest paths. The hop-d... |

1029 | Multicast routing in a datagram internetwork - DEERING - 1991 |

955 | The Art of Computer Systems Performance Analysis - Jain - 1991 |

693 |
Routing of multipoint connections
- Waxman
- 1988
(Show Context)
Citation Context .... Donahoo are with the College of Computing, Georgia Institute of Technology. 1 \Well-known" topologies, such as the ARPAnet or NSFnet backbone (e.g., [5, 54,65]) Randomly generated topologies (e.g., =-=[55, 58, 61]-=-) The limitations of each of these are obvious: well-known and regular topologies re ect only parts of current or past real networks; random topologies may not re ect any (past, present or future) rea... |

660 |
Design and Analysis of Experiments
- Montgomery
- 2001
(Show Context)
Citation Context ... means of either at method compared to the Transit-Stub method are not equivalent. Note that our large sample size (1000 graphs/runs) relaxes the need for normality in this type of hypothesis testing =-=[45]-=-. Figure 11 shows the distribution of multicast performance for each metric. Note that, as expected, the frequency distributions for the two at methods are very similar, and the Transit-Stub frequency... |

539 | Supporting real-time applications in an integrated services packet networks: architecture and mechanism - Clark, Shenker, et al. - 1992 |

536 | TCP Vegas: New techniques for congestion detection and avoidance
- Brakmo, O'Malley, et al.
- 1994
(Show Context)
Citation Context ...es of how the results scale or how they might change with a di erent topology are so rare. The state of the art in network modeling includes: Regular topologies, such as rings, trees and stars (e.g., =-=[12, 44, 64]-=-) E. Zegura, K. Calvert, and M. Donahoo are with the College of Computing, Georgia Institute of Technology. 1 \Well-known" topologies, such as the ARPAnet or NSFnet backbone (e.g., [5, 54,65]) Randoml... |

504 | The PIM Architecture for Wide-Area Multicast Routing - Deering, Estrin, et al. - 1996 |

451 | Modeling Internet topology
- Calvert, Doar, et al.
- 1997
(Show Context)
Citation Context ...enerated are all connected, this construction results in a connected graph. Doar has independently proposed a graph generation method that has a somewhat similar hierarchical structure to this method =-=[15]-=-. The size of the graph (number of nodes) and the distribution of nodes between transit and stub domains in this method is controlled by the following parameters: 5 Param. Meaning Ex. T number of tran... |

440 | Distance Vector Multicast Routing Protocol
- Waitzman, Partridge, et al.
- 1988
(Show Context)
Citation Context ...owing comparisons across instances with di erent numbers of sources and receivers. In addition, shortest-path trees are the basis for the presently-deployed Internet multicast routing algorithm DVMRP =-=[25, 48]-=-, thus the ratio allows an evaluation of how algorithm A compares to an accepted standard.sDelay ratio. This metric considers the maximum delay, measured in hops, from any source to any receiver in th... |

259 | A fast algorithm for Steiner trees - Kou, Markowsky, et al. - 1981 |

226 | Firewalls and Internet Security: Repelling the Wily Hacker - Cheswick, Bellovin - 1994 |

203 | Multicast routing for multimedia communication - Kompella, Pasquale, et al. - 1993 |

158 | The Stanford GraphBase: A Platform for Combinatorial Computing - Knuth - 1993 |

141 | Dynamic Steiner tree problem - Imase, Waxman - 1991 |

137 | A tradeoff between space and efficiency for routing tables - Peleg, Upfal - 1989 |

130 |
Empirically derived analytic models of wide-area TCP connections
- Paxson
- 1994
(Show Context)
Citation Context ... according to the KS test. N/A entries indicate that we did not perform experiments for this particular graph size and node degree. 9 Other methods exist for comparison of sample distribution (e.g. 2 =-=[49]-=-). We chose to use a method that is somewhat more simple to evaluate and is not sensitive to factors such as selection of bin size. 10 Clearly some information is lost by this representation, however ... |

130 | The trade-offs of Multicast Trees and Algorithms
- Wei, Estrin
- 1994
(Show Context)
Citation Context ...th trees for a network model with average node degree of about 3.0, but the tra c concentration is almost 30% higher in core-based trees when average node degree increased to 8.0. (See Figure 9(a) in =-=[63]-=-.) Mitzel and Shenker found that multicast resource reservation styles compared quite di erently in the quantity of resources reserved for linear, tree and star topologies. (See Tables 4 and 5 in [44]... |

123 | Core based trees - Ballardie, Francis, et al. - 1993 |

121 | How Bad is Naïve Multicast Routing
- Doar, Leslie
- 1993
(Show Context)
Citation Context ...ting algorithms was reduced byasmuch as a half when using random graphs versus using hierarchically structured graphs designed to re ect some of the properties of real internetworks. (See Figure 5 in =-=[26]-=-.) Wei and Estrin found that the tra c concentration in core-based multicast routing trees is comparable to tra c concentration in shortest-path trees for a network model with average node degree of a... |

101 | Local area network traffic characteristics with implications for broadband network congestion management - Fowler, Leland - 1991 |

95 | Routing to Multiple Destinations in Computer Networks - Bharath-Kumar, Jaffe - 1983 |

66 | Mechanisms for Broadcasts and Selective Broadcasts - Wall, in - 1980 |

65 | Balancing minimum spanning and shortest path trees - Khuller, Raghavachari, et al. - 1993 |

63 | Wide-area tra c: The failure of poisson modeling - Paxson, Floyd - 1995 |

44 | Core Selection Methods for Multicast Routing
- Calvert, Zegura
- 1995
(Show Context)
Citation Context ...t? To determine this e ect, we consider the problem of multicast routing. We assume the reader is familiar with the concepts of multicasting, and center-based (or shared-tree) multicast routing. (See =-=[16]-=- for details). Many di erent generation methods have been used to study multicast routing algorithms, including the Waxman method [61] and well-known topologies such as the ARPAnet [63]. Thus, it is i... |

42 | a Network Simulator - “REAL - 1988 |

40 | On the Self-Similar Nature of Ethernet Tra c - Leland, Taqqu, et al. - 1993 |

36 | The maximum connectivity of a graph - HARARY - 1962 |

33 | On finding Steiner vertices - Rayward-Smith, Clare - 1986 |

32 |
Digital Telephony
- Bellamy
- 2000
(Show Context)
Citation Context ...ect the current topology. Historically, large networks such as the Public SwitchedsTelephone Network have grown according to a topological design developed by some central authority or administration =-=[7]-=-. In contrast, there is no central administration that controls |or even keeps track of| the detailed topology of the Internet. Although general characteristics of its topology are known, it is imprac... |

32 | The network simulator - Heybey - 1990 |

30 | Performance Evaluation of Multipoint Routing Algorithms - Waxman - 1993 |

29 | MOSPF: Analysis and Experience - Moy - 1994 |

29 | Controlling Alternate Routing in General-Mesh Packet Flow Networks
- Sibal, Desimone
- 1995
(Show Context)
Citation Context ...(e.g., [12, 44, 64]) E. Zegura, K. Calvert, and M. Donahoo are with the College of Computing, Georgia Institute of Technology. 1 \Well-known" topologies, such as the ARPAnet or NSFnet backbone (e.g., =-=[5, 54,65]-=-) Randomly generated topologies (e.g., [55, 58, 61]) The limitations of each of these are obvious: well-known and regular topologies re ect only parts of current or past real networks; random topologi... |

21 | Protocol independent multicast (PIM): Motivation and architecture, Internet draft - Deering, Estrin, et al. - 1995 |

21 | Estimates of Distributions of Random Variables for Certain Computer Communications Traffic Models - Fuchs, Jackson - 1970 |

19 | Protocol independent multicast (PIM), sparse mode protocol: Specification - Deering, Estrin, et al. - 1996 |

18 | Asymptotic resource consumption in multicast reservation styles
- Mitzel, Shenker
- 1994
(Show Context)
Citation Context ...es of how the results scale or how they might change with a di erent topology are so rare. The state of the art in network modeling includes: Regular topologies, such as rings, trees and stars (e.g., =-=[12, 44, 64]-=-) E. Zegura, K. Calvert, and M. Donahoo are with the College of Computing, Georgia Institute of Technology. 1 \Well-known" topologies, such as the ARPAnet or NSFnet backbone (e.g., [5, 54,65]) Randoml... |

15 |
Routing reserved bandwidth multi-point connections
- Verma, Gopal
- 1985
(Show Context)
Citation Context .... Donahoo are with the College of Computing, Georgia Institute of Technology. 1 \Well-known" topologies, such as the ARPAnet or NSFnet backbone (e.g., [5, 54,65]) Randomly generated topologies (e.g., =-=[55, 58, 61]-=-) The limitations of each of these are obvious: well-known and regular topologies re ect only parts of current or past real networks; random topologies may not re ect any (past, present or future) rea... |

13 | Host extensions for IP multicasting,” Internet Request For Comments - Deering - 1989 |

12 |
Optimizing File Transfer Response Time Using Loss-Load Curves: A Simulation Study
- Williamson
- 1995
(Show Context)
Citation Context ... how the results scale or how they might change with a different topology are so rare. The state of the art in network modeling includes: ffl Regular topologies, such as rings, trees and stars (e.g., =-=[12, 44, 64]) E. Zegur-=-a, K. Calvert, and M. Donahoo are with the College of Computing, Georgia Institute of Technology. ffl "Well-known" topologies, such as the ARPAnet or NSFnet backbone (e.g., [5, 54, 65]) ffl ... |

10 | A comparison of two practical multicast routing schemes - Calvert, Madhavan, et al. - 1994 |

9 | Low cost algorithms for message delivery in dynamic multicast groups - Belkeir, Ahamad - 1989 |

8 |
Policy Routing in Internet Protocols. Internet Engineering Task Force
- CLARK
- 1989
(Show Context)
Citation Context ...n can be easily included as an overlay to the basic topological structure, as needed for studying particular problems. Today's Internet can be viewed as a collection of interconnected routing domains =-=[19]-=-, which are groups of nodes that are under a common administration and share routing information. A primary characteristic of these domains is routing locality: the path between any two nodes in a dom... |

8 | Local Area Network Tra c Characteristics with Implications for Broadband Network Congestion Managment - Fowler, Leland - 1991 |