## Scaling of Multicast Trees: Comments on the Chuang-Sirbu Scaling Law (1999)

### Cached

### Download Links

- [www.cs.umd.edu]
- [www.isi.edu]
- [ftp.isi.edu]
- DBLP

### Other Repositories/Bibliography

Venue: | In Proceedings of ACM SIGCOMM ’99 |

Citations: | 111 - 5 self |

### BibTeX

@INPROCEEDINGS{Phillips99scalingof,

author = {Graham Phillips and Scott Shenker and Hongsuda Tangmunarunkit},

title = {Scaling of Multicast Trees: Comments on the Chuang-Sirbu Scaling Law},

booktitle = {In Proceedings of ACM SIGCOMM ’99},

year = {1999},

pages = {41--51}

}

### Years of Citing Articles

### OpenURL

### Abstract

One of the many benefits of multicast, when compared to traditional unicast, is that multicast reduces the overall network load. While the importance of multicast is beyond dispute, there have been surprisingly few attempts to quantify multicast’s reduction in overall network load. The only substantial and quantitative effort we are aware of is that of Chuang and Sirbu [3]. They calculate the number of links L in a multicast delivery tree connecting a random source to m random and distinct network sites; extensive simulations over a range of networks suggest that L(m) 0: rr~‘.~. In this paper we examine the function L(m) in more detail and derive the asymptotic form for L(m) in k-ary trees. These results suggest one possible explanation for the universality of the Chuang-Sirbu scaling behavior. 1

### Citations

1317 | On power-law relationships of the Internet topology
- Faloutsos, Faloutsos, et al.
- 1999
(Show Context)
Citation Context ...he predicted form applies less well. Our original intuition that real networks have exponential 6 However, the existence of exponential growth in real networks is the subject of some controversy; see =-=[8]-=- for a different reading of similar data. ts1008 ti5000 ts1000 r100 r T(r) 35 30 25 20 15 10 5 0 10000 1000 100 10 1 AS Internet MBONE ARPA r 30 25 20 15 10 5 0 100000 10000 1000 100 10 1 (a) generate... |

1020 | Multicast routing in datagram internetworks and extended LANs
- Deering, Cheriton
- 1990
(Show Context)
Citation Context ... asymptotic form for L(m) in k-ary trees. These results suggest one possible explanation for the universality of the Chuang-Sirbu scaling behavior. 1 Introduction IP multicast routing was proposed in =-=[4, 5, 6]-=-. While there have been frustrating delays in its widespread commercial deployment, experience with multicast in the MBone [2] and other experimental settings has provided convincing evidence of multi... |

685 |
Routing of Multipoint Connections
- Waxman
- 1988
(Show Context)
Citation Context ...ntial reachability functions. The question then becomes: do real networks (current or future ones) have exponential reachability functions S(r)? More investigations of artificially generated networks =-=[1, 7, 11, 10]-=- and real networks (see [9, 8]) would shed light on this question, and this will be the subject for our future research. If exponential behavior in S(r) is generic then we think our analysis helps exp... |

445 | Modeling Internet Topology
- Calvert, Doar, et al.
- 1997
(Show Context)
Citation Context ...g connectivity between autonomous systems collected by the National Laboratory for Applied Network Research (NLANR). 2 The rest of the networks are generated either using the GT-ITM network generator =-=[1] or the TIERS networ-=-k generator [7]. The names of the generated graphs reveal their topological style: "ts" refers to transit-stub networks (generated by the GT-ITM network generator), "r" refers to r... |

284 | Multicast routing in internetworks and extended LANs - Deering - 1988 |

166 | A better model for generating test networks
- Doar
- 1996
(Show Context)
Citation Context ...ystems collected by the National Laboratory for Applied Network Research (NLANR). 2 The rest of the networks are generated either using the GT-ITM network generator [1] or the TIERS network generator =-=[7]. The names of the g-=-enerated graphs reveal their topological style: "ts" refers to transit-stub networks (generated by the GT-ITM network generator), "r" refers to random networks (generated by the GT... |

128 | The Tradeoffs of Multicast Trees and Algorithms
- Wei, Estrin
- 1995
(Show Context)
Citation Context ... multicast routing algorithms that are source specific, where packets traverse the shortest path between source and receiver; we do not address the efficiency of shared tree multicast algorithms. See =-=[12]-=- for one such comparison. name type source nodes links avg. degree ARPA real Early ARPANET 47 68 2.9 MBone real MBone (Feb 1999) 4179 8529 4.1 Internet real partial Internet (March 1999) 56317 77154 2... |

113 | First IETF Internet audiocast
- Casner, Deering
- 1992
(Show Context)
Citation Context ...aling behavior. 1 Introduction IP multicast routing was proposed in [4, 5, 6]. While there have been frustrating delays in its widespread commercial deployment, experience with multicast in the MBone =-=[2]-=- and other experimental settings has provided convincing evidence of multicast's power and elegance. From a conceptual viewpoint, multicast offers applications a novel logical rendezvous mechanism whe... |

72 | Pricing Multicast Communications: A Cost-Based Approach
- Chuang, Sirbu
- 1998
(Show Context)
Citation Context ... dispute, there have been surprisingly few attempts to quantify multicast's reduction in overall network load. The only substantial and quantitative effort we are aware of is that of Chuang and Sirbu =-=[3]-=-. They calculate the number of links L in a multicast delivery tree connecting a random source to m random and distinct network sites; extensive simulations over a range of networks suggest that L(m) ... |

39 |
Discovering Internet Topology
- Siamwalla, Sharma, et al.
- 1998
(Show Context)
Citation Context ... less well to the MBone. The MBone remains partially an overlay network, which may affect the nature of T (r). The nature of S(r) for real networks is an interesting, and open, research question: see =-=[9]-=- and [8] for some related work on this topic. 4.3 Non-exponential Cases While we expect that S(r)se r for somesfor many networks, we have already seen that there are cases where the reachability funct... |

26 | A comparison of multicast trees and algorithms
- Wei, Estrin
- 1994
(Show Context)
Citation Context ... multicast routing algorithms that are source specific, where packets traverse the shortest path between source and receiver; we do not address the efficiency of shared tree multicast algorithms. See =-=[12]-=- for one such comparison. name type source nodes links avg. degree ARPA real Early ARPANET 47 68 2.9 MBone real MBone (Feb 1999) 4179 8529 4.1 Internet real partial Internet (March 1999) 56317 77154 2... |

2 |
Models for multipoint connections in gigabit networks
- Waxman
- 1995
(Show Context)
Citation Context ...ntial reachability functions. The question then becomes: do real networks (current or future ones) have exponential reachability functions S(r)? More investigations of artificially generated networks =-=[1, 7, 11, 10]-=- and real networks (see [9, 8]) would shed light on this question, and this will be the subject for our future research. If exponential behavior in S(r) is generic then we think our analysis helps exp... |