GLOCOM Review 8:4 (76-1) 2003 Center for Global Communications 1
2
GLOCOM Review 8:4 (76-1) Recent research has been providing more support for the idea that Internet layers are formed along Power Law lines. Network systems, which follow Power Law principles, show scale-freeness and unequal distributions, and examining them can deepen our understanding of the Internet itself. The present paper focuses on the most basic levels of the Internet backbone, to determine whether or not it's network structure is governed by the Power Law. The analysis is based on data for IIJ, which is believed to have parameters representative of domestic information systems. IIJ is a rapidly expanding network, and it follows the Power Law, which can be inferred to be the case for the Internet backbone as a whole. Taking a closer look at IIJ's growth processes revels that following the geographic stabilization of NOCs and the rapid increase in circuit volume in 1998, there was a numerical increase in Power Law Indexes (Zipf Indexes) and also an increasing concentration of site hubs in Tokyo. It is true that the early years of 1994-97 do not show stabilized Power Law distribution, and 1995, the year in which the Telehodai system went into effect, shows untypical figures. On the whole, however, applying Power Law Indexes to network analysis has shed light on the network structure of and growth process of backbone circuiting. Keywords; Internet backbone circuiting, Power Laws, Power Law Indexes, Tokyo-based centralization, Regional IX, IIJ, "hubs" 3
1993 1 1997Web 2 1999WWW 3 4 IXInternet exchange 4
5 GLOCOM Review 8:4 (76-1) 1 2 2
5 6 A 7 2000 8 2 y x 1 y = bx 1 xr2 1 x = ar 2 2 x R 3 R = a 1/ x 1/ 3 N r 3 r = ( A / N ) x 1/ A = a 1/ 3 6
GLOCOM Review 8:4 (76-1) x p 3 4 x p [] = 1( A / N ) x 1/ 4 45 1 p ' ( A / N) x 1/ 5 11 log y = log blog x 1 x 3 7
9 8020 28 2002 12320641 78.92 88020 10 2 11 2 2 IIJ IIJ 8
GLOCOM Review 8:4 (76-1) 21NOC A-Bone IIJ ISP 12 NOCNOC NOCNOC NOCNOC NOC NOC NOC 2002121NOC21 21NOC 4,800Mbps 2,700Mbps1,200Mbps 4 TREE2 33 12 2 9
NOC 62002IIJNOC 26,763Mbps9,692Mbps2,700Mbps 2,001MbpsNOC NOC NOCNOC 20.91 NOC 2.66 10
GLOCOM Review 8:4 (76-1) 11
19942002IIJ 719942002NOC1994 4 NOC19958 NOC200121 NOC8517 NOC1996 3NOC1999453 20015 NOC IIJ 919946.7Mbps200247,866Mbps 87 2.3 IIJ199420027 NOC NOC 19942002 81995 2.12.7 12
GLOCOM Review 8:4 (76-1) 13
9 1998 199820024 0.5 1998 199497 19951.611994972.2 1995 238 19958NTT 1994 NetscapeWebWindows 95TCP/IP PC 19971998 19977JPIXKDDMEXIIJ OCNODNDIONISP Bekkoame RIMNETISP IIJIIJ 14
GLOCOM Review 8:4 (76-1) 12 IIJ199420027 NOC 1NOC 2 19942002NOC 1998 199497 1995 NetscapeWebWindows 95 TCP/IPPC 199798 1997JPIXKDDMEXIIJ ISPBekkoameISP 15
1 M.Takayasu,and H.Takayasu [1993] "Fractals 1." 371 2 NTT1997Web Web 3 =2002NHK pp.96-115 4 2002IX 5 Auerbach.F [1913] "Das Gesets der Bevolkerungsknzentration." Petermann's Mitteilungen, 59, pp.74-76 6 <http://linkage.rockefeller.edu/wli/zipf/> 7 Lada A. Adamic "Zipf, Power-laws, and Pareto - a ranking tutorial" Xerox Palo Alto Research Center <http://ginger.hpl.hp.com/shl/papers/ranking/ranking.html> 8 2003 9 x x R R = Ax 10 199818 11 =2002NHK 12 2003.4 16