Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
1
LawsThat Are GoverningThe Network
Harvey M. GatesInterdisciplinary Telecommunications Program
School of Engineering and Applied Sciences
Engineering Center OT 3-17
Campus Box 530
University of Colorado at BoulderBoulder, CO 80309-0530
303-735-2684
2000 Third Annual International Symposium on Advanced Radio Technologies
September 6-8, 2000
U.S. Department of Commerce
Boulder Laboratories
Boulder, CO
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
2
THE NETWORK
• “The Network”– All digital networks worldwide– Includes the Internet– Includes the digital carrier networks
• Local Exchange Carriers• Inter-Exchange Carriers
– Includes trans-oceanic carriers– Includes all digital networks
• Local area networks (LANs)• Metro area networks (MANs)• Wide area networks (WANs)
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
3
LAWS OF THE NETWORK
• Gordon Moore’s Laws– Moore’s 1st Law– Moore’s 2nd Law
• George Gilder’s Law of the Telecosm• Network Intelligence
– Peter J. Sevcik on Switched System Performance
– David Isenberg and the Rise of the Stupid Network
• Observations in Satellite Systems• Disruptive Technologies, a Need• Conclusions
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
4
MOORE’S 1st LAW
• Moore’s Law (The 1st Law)– Gordon Moore, Chairman Emeritus of Intel, predicted
that chip complexity would double every device generation
• Chip complexity is defined by the number of active elements on a single semiconductor chip
– Now roughly comparable to performance as measured in millions of instructions per second (MIPS)
• The device generation was assumed to be approximately 18 months, and still applies today
• Valid now for three decades
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
5
MOORE’S 1st LAW PREVAILS
• Appears likely to be valid for several more device generations
1975 1997 2003Chip complexity (index to 1) 1 10 100Feature size reduction, mm 2 0.25 0.08Chip size increase, mm2 30 150 600Wafer diameter, mm 50 200 300Facility automation, % 5 60 80Die yield, % good 40 85 95Line yield, % good 40 90 95Operational efficiency 1 10 100Equipment cost 1 10 50
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
6
• Cost in fabricating chips (factory cost) is also exponential in the opposite direction and is off-setting the gains in complexity
– High-volume factories are now multi-billion dollar investments
MOORE’S 2nd LAW
101
102
103
104
105
106
107
108
109
Com
plex
ity
1960 1965 1970 1975 1980 1985 1990 1995 2000
$200
$500
$2000
$5000
Fact
ory
Cos
t in
$M
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
7
INERTIA BEHIND MOORE
• Side note:– In 1974, Moore extrapolated the wafer size suggesting that by
2000 a wafer would be 57-inches in diameter - off by a factor of 10
• Important to note the impact of new technologies, materials, and processes allow structures on an atomic scale– New products and micromechanical devices in the labs include
• Micro-refrigerators• Micro-turbines• Micro-motors
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
8
GILDER’S LAW OF THE TELECOSM
• The law of the telecosm ordains that the total bandwidth of communications systems will triple every year for the next 25 years - George Gilder, Gilder Technology Report Volume II, Number 2, February 1997
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
9
IMPRESSIVE FIBER GROWTH19
86
1988
1990
1992
1994
1996
Fiber Deployment by Telcos
20
18
16
14
12
10
8
6
4
2
0
Mill
ions
of F
iber
Mile
s
CLECs
ILECs
IXCs
Source: FCC
Global Undersea Fiber Capacity
300
250
200
150
100
50
0
Gig
abits
Source: GTG19
84
1986
1988
1990
1992
1994
1996
1998
2000
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
10
ADVANCES IN WAVE DIVISION MULTIPLEXING (WDM)
• WDM commercially introduced in 1996• Now running 8, 16, and 32 λs• Avanex now testing 800• Lucent’s “AllWave product objective is 3300 λs
800 1000 1200 1400 1600 1800
Wavelength (nm)
25
Single modeMultimode
105
21
.5
.2.1A
ttenu
atio
n (d
B/k
m)
Visible light is between 400 to 700 nm.
3 THz
WDM channels
20
15
10
5 Opt
ical
am
p ga
in, d
B
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
11
MORE BANDWIDTH ON THE WAY
• Bundled fibers• Moves toward lower noise & higher amplification
efficiencies– Pre-Erbium doped fiber amplifiers (EDFAs) repeater
spacing on the order of less than 100 km– EDFAs allow 400 to 600 km– Ramon amplifiers show promise of 10 Gb/s to 3,000 km– Ytterbium is another doping possibility
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
12
BUNDLED FIBER OPTIC TRANSMISSION SYSTEMS
27
32
• Bundle of 864 SMF strands each with a core, cladding and buffer
• At 3,300 λs each• Could contain 2.86 million λs• At 10 Gigabits/ λ, that equals
28.6 Petabits/sec for this bundle
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
13
NETWORK INTELLIGENCE
• Network intelligence is moving to the network edge and away from the network core
– David S. Isenberg in the Rise of the Stupid Network• Concept introduced April 1, 1996• Followed by internal AT&T publications and seminars
– Peter J. Sevcik’s Network Switching Laws were published a year later in the Business Communications Review, September, 1997
• Both weave the transition from centralized to decentralized network intelligence
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
14
DAVID S. ISENBERG’S MESSAGES
• On April 1, 1996, Isenberg, a top scientific researcher at AT&T,sponsored an international conference titled: What if minutes Were Free?
– So heretical and ridiculous that he chaired the event in a court jester’s outfit.
• An internal paper on the subject was circulated at AT&T• A year later, June 25, 1997 at 5:48 PM EST, Isenberg was allowed to
release the electronic version - Rise of the Stupid Network– He wrote in his cover e-mail: “Here is my attempt at entropy gradient
reversal at AT&T. . .– . . If you want EXPLICIT frontal exposure of the Critical Issues, or
language that is commensurate with my thinking, you’ll have to read between the bits.”
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
15
ISENBERG’S SIMPLE PREMICE - THE MIPS OF A TELEPHONE VERSES THE PC AND END USER IS SHIFTING
POWER TO THE EDGE OF THE NETWORK
4
44
4 4
IXCNetwork
OfFiber
TrunksThe
AssetsOf the
IXC
ISPRouters
Client
Server
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
16
SEVCIK'S NETWORK SWITCHING LAWS
• Sevcik published his work in the Business Communications Review,September, 1997, page 33, titled “Why Circuit Switching is Doomed”
– Short and sweet quantitative material that fell in line with Isenberg’s findings• Sevcik’s Network Switching Laws
– Successful new switching technologies double their performance/cost ratio twice as fast as the previous technology.
– As switches improve their performance/cost ratios, they also shed processing functions to satellite processors (severs) or directly to end stations.
– A steep performance/cost improvement trajectory is more important to the success of a switching technology than its initial performance or cost relative to its competitors.
• Side note: “Personally, I never bet against compounding.” Peter Sevcik
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
17
MOST INTERESTING GRAPH
Peter J. Sevcik"Why Circuit Switching is Doomed"Business Communications Review, September, 1997
10,000,000
1,000,000
100,000
10,000
1,000
100
10
1980 1985 1990 1995 2000
Perf
orm
ance
/Cos
t (bp
s per
$)
Circuit Switch Router Switch
Fram
e Switc
h
ATM Switch
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
18
ELECTRONIC VERSES OPTICAL SWITCHING
• More support for dumbing down networks– Micro Electro Mechanical Systems (MEMS)
• Lucent Technologies’ WaveStar™ LambdaRouter™ of 256-inputs to 256-outputs
– Bubble Matrix• Agelent Technologies’ Champagne
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
19
ARE SATELLITES KEEPING PACE?
• Research conducted by Victoria L. Miralda at the University of Colorado - Spring Semester, 2000
– Satellites may be an aberration in the data– Some data - commercial payloads deployed
GEO Commercial LEO Commercial Total CommercialPayloads
1995 18 4 221996 24 1 251997 28 59 871998 22 82 1041999 22 54 76
AST: Commercial Space Transportation: 1999 Year in Review
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
20
SATELLITE LAUNCHES SEEM TO DOUBLE EVERY TEN YEARS
13
73
171247
1960 1970 1980 1990 2000
250
200
150
100
50
0Num
ber
of G
EOs L
aunc
hed
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
21
ROLL-UP OF RAW ORBITAL BANDWIDTH CURRENT AND PROJECTED
1500
1000
500
01990 1995 2000 2005 2010
80.04 159.1
688.5
1145
Band
wid
th in
GH
z
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
22
LAW OF THE SPACECOSM
• LEOs will be capable of tripling capacity approximately every 5 years and GEOs every 8-10, depending on the multiple factors of the Spacecosm. The average time to triple satellite bandwidth capacity will be on average 7 years.
Victoria L. Miralda, “Will Satellite Communications Keep Pace With Internet?” Research Thesis, University of Colorado at Boulder, Interdisciplinary Telecommunications Program, Boulder, CO, April 14, 2000
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
23
A CASE FOR DISRUPTIVE TECHNOLOGIES
• A disruptive technology is one which defies projections by introducing a step function in the trend that it supports– Marc Andreessen’s Mosaic Browser given Ted Nelson’s hyper-
text and Tim Berners-Lee hyper-linking– Breakthrough of the semiconductor industry from traditional Al+Si
vrs CU+Cu– Wave Division Multiplexing (WDM)– And so on . . .
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
24
CLIENTS AND SERVER DEMANDS WILL NOT SLOW DOWN
Digitally downloaded product sales forecast, 1999 to 2004
$3,000M
$2,500M
$2,000M
$1,500M
$1,000M
$500M
0
1999 2000 2001 2002 2003 2004
VideosBooksVideogames
Music
Software
September 7, 2000University of Colorado at BoulderInterdisciplinary Telecommunications Program
25
CONCLUSIONS
• The laws of the network seem to be holding firm• Such projections have and will continue to appear radical and even
preposterous– “Here is my attempt at entropy gradient reversal at AT&T. Of course,
everyone knows that reversing the entropy gradient is absolutelyimpossible, and that if you show even the vaguest threat of succeeding, the threatened world throws you out on your ear.” - David S. Isenberg
• But, disruptive technologies can be counted on - the final law of the network
• Let each and every one of us be a part of this wonderful ride