!@#$%^&*()!@#$%^&*()!@#$%^&*()!@#$%^&*()!@#$%^&*()!@#$%^&*()
------------------------------------------------------------
SMART Letter #5 -- April 14, 1998
For Friends and Enemies of the Stupid Network
Copyright 1998 by David S. Isenberg
This document may be redistributed provided that
the 11 lines containing this notice accompany it.
isen@isen.com -- http://www.isen.com/ -- 1-888-isen-com
It takes SMART people to design a Stupid Network
------------------------------------------------------------
!@#$%^&*()!@#$%^&*()!@#$%^&*()!@#$%^&*()!@#$%^&*()!@#$%^&*()
LOOKING BEYOND TODAY'S STYLES
These days, you can't be too rich, or too young, or have too much
dark fiber buried down by the railroad track. The new bandwidth
mongers - Qwest and Level3 and Williams and Worldcom's Resurgens
- are the belles of tonight's ball. Will the FCC spoil the party?
Saturday's New York Times says, "Yes." Monday's Wall Street Journal
says, "Not just yet." The White House and the FCC are on opposite
sides, and the industry is clearly split between the old Intelligent
Network crowd and the new Stupids.
Regulation is the biggest uncertainty, but technology marches on.
Struggling against technological advance, especially when it is in
the service of fundamental human needs, seems distracting and
wasteful. If the captains of the great age of sail could've taxed
the steam ship, they would have. Wouldn't have changed a thing.
At the Last Mile, the surging, synergistic pseudo-conflict between
DSL and Cable Modems is on every investor's lips. The winners?
Probably Bill Gates, Rupert Murdoch, Mike Eisner, Barry Diller,
and The Dreamworks Three. And L.L. Bean and Amazon.com and Fedex.
Oh, yeah . . . you (the SMART telecom provider) and me (the consumer).
SMART people know that you can't predict the future. Yesterday's
sure thing - Apple Computer, Japanese management, a 500 channel
teleTV utopia - is tomorrow's tailfinned has-been.
Big Bandwidth is coming. But let's try not to overestimate
short-term impact, or underestimate long-term impact.
Below we ask what is driving the demand, and how the
supply story will evolve.
BANDWIDTH SKEPTICISM: At least one SMART person, David Charlton,
is a blatant bandwidth skeptic, openly puzzling over where all the
traffic growth is coming from, and whether it will continue.
Skepticism, yes. Dismissiveness, no. Where IS it all coming from?
Coming it is. Mike O'Dell, UUNet's CTO, is adding a DS-3 per day
to the UUNetwork. He has to, because his bandwidth providers need
45 to 90 days of lead time. And his network, which merely doubled
in 1994, grew by a factor of 10 in 1997. "Only epidemics grow this
fast," says O'Dell.
The virulence is contagious. Vint Cerf, MCI's
Internet Pointman, has said that MCI's data traffic is doubling
twice a year. And Scott Gorman, sales manager for Exodus
Communications (and one of the SMART), says that the aggregate
throughput of Exodus Internet Data Centers has grown by a factor
of 40 in the last 12 months. Exodus has adopted a 200% headroom
policy - a good example of Stupid Network Principle #2,
Overprovisioning. That is, Exodus buys twice as much bandwidth
as it needs. And keeps doing it, cause it doesn't sit on
the shelf long at all!
Undeniably, the growth is there, and accelerating, no matter how
much scalp we scratch. "If you are not scared, you don't understand,"
says Mike O'Dell. Well, Mike, we're scared, and *we*still*don't*
*understand*. Where is it coming from? Like, almost nobody has
anything faster than 56 kbit/s yet. Like, most PC sales are
replacements, not new users. Like *what* new killer apps?
SUPPLY CLUB: Another SMART person, Gordon Cook, does see bandwidth
demand skyrocketing, but thinks that George Gilder and his buddies
(moi?) are all "starry eyed," because there's "not one shred of
evidence" that "vastly growing supply [of bandwidth] will mean the
price will plummet."
Well here's a shred. Average per-minute LD
revenue has dropped from 32 cents in 1984 to 16 cents in 1996.
OK, it is a pretty pitiful shred.
Here's a bigger one - today's global Internet itself.
In 1956, Ma Bell paid $5.30 per hour for a trans-Atlantic
voice circuit - there were 89 of them. In those days, would
UUNet have let you suck over hour after hour of BBC via
Real Audio? The Internet only works as it does today
because the cost of bandwidth is negligible. These costs dropped
by a factor of 10 over the 15 years 1956-1970, and another factor
of 10 over the next 15, so in 1993 an hour of voice circuit cost
3 cents. (See the graph at
http://www.itu.int/ti/wtdr95/graphics/ov6.gif).
One could look at these numbers and ask why an international
call that costs three cents an hour is priced at around US$30.
To compare, let's look at a few more mark-ups. I'd guess there's
about three cents of sugar and bubbles in a can of Coke, yet we
don't pay US$30 for it. On the other hand, there's about - what,
US$3.00? - worth of chemicals in the human body, and the mark-up,
measured by wrongful death settlements, is a few hundred thousand
times. This mark-up would have you paying US$3000 for an hour-long
three-cent-cost phone call.
But let's not mistake a short distance for a clear view. Trans-Atlantic
calls are now down to 10 cents a minute (US to UK, off peak) and
falling. European de-reg is just kicking in, and the WTO telecom
agreement has not yet been felt. Trans-Oceanic capacity is poised
for another growth spurt, and the cable-opoly knows it won't make
money if it does not fill its pipes.
On the other hand, the members of this hypothetical cable club
would be shooting themselves in the demand side if they drop the
prices so fast that their new pipes are full on the first day.
They aim to optimize return on their investment. Don't we all?
In non-carrier sectors, prices are clearly plummeting. A 10baseT
socket in your pocket costs a few bucks. You get 1000 chunks of
64 kbit throughput for US$1.00 when you buy an off-the-shelf Gigabit
Ethernet platform. (When's the last time you paid to log into a
LAN?) And my cable megabidem costs about what my 300 baud Hayes
did 20 years ago. No cozy club is damping these price declines.
There are too many players for that.
New carriers are in a bind.
They have to drop their prices to gain market share, and they want
to join the club. But the club is exclusive, and it won't admit
price-droppers. Price wars are soooo unGENtlemanly. Meanwhile
the current members are busy uping their capacity too. New and
old chase the same commodity bandwidth market.
These are ingredients for instability.
Even with stable, growing markets, the price can
only go one way. It's Econ 101. Qwest's Joe Nacchio says that he
chose a 7.5 cent price for voice because he can offer it at the
same (ahem, comfortable) profit margins at which the big guys offer
full priced voice. There is room for that price to fall. If the
supply outruns the demand, Nacchio's prices are likely to come down
as fast as necessary.
Let's look at another way you can play the supply-demand game with
improving technology. For years, Intel and Microsoft played
scratch-my-back - the software got more complex as the chips became
more capable. The price-per-box hovered at $3000. Until this
year, when people discovered that they did not need faster computers
to run the current software. Then the price fell. (And now
Microsoft's and Intel's market will expand in new, non-PC-oriented
ways.) If I were sitting on a pile of new bandwidth, I'd want to
find more things for people to do with it, and drop prices just
enough to make those things happen.
LOCAL BOTTLENECKS: One might observe recent moves by AOL, IBM and
AT&T, and conclude that the cost of Internet connections is going
up, not down. But let's not confuse access with the price/cost of
long haul bandwidth. Circuit access ties up two modems and a
twisted pair, whether or not there is traffic. It is a hold-over,
a scarcity-based design.
IBM and AT&T have ended their US$19.95/month unlimited connect-time
plans. They have placed limits on the number of hours of use per
month. "If you ain't using, hang up." IBM seems to have installed
an inactive-logoff program to enforce this.
But IBM and AT&T are getting it wrong. Here's what they're missing.
>From a user's standpoint, there's value in being logged on, even
if inactive. The barrier to using network services (e.g., checking
email) is lower. You can get an Internet phone call. You can know
if a buddy-list buddy logs on. It is a shorter step to impulse
buying. This overly intelligent move by IBM and AT&T will send
traffic, customers, and revenues to other providers. These days a
big installed base is worth a lot more than a big profit margin.
AOL has a better handle on the circuit access situation.
They have raised the all-you-can-eat price from $19.95 to $21.95.
They hardly lost anybody, and the small proportion of customers
who have discovered the need for always-on can still get it
painlessly. AOL's market segment will wind up becoming more
desirable versus AT&T's and IBM's.
ALWAYS-ON, THE BUS: Ultimately, we want to get rid of circuit
access in favor of bus-style access. Last night, a conversation
with dinner guests hinged on the price of an Audi A4. I seamlessly
slid behind my always-on, Comcast Cable Modem powered Internet
terminal. Metacrawler . . . boom. "Automobile Price Audi A4," .
. . boom, boom. And there it was - list price, dealer invoice,
options. Logging on probably would have been a big enough barrier
to have kept me from doing it.
Bus-architected systems, e.g., cable access, are designed to be
always-on. Now, for the first time, with my cable modem, when I
click into the audio stream of an Internet radio station,
*I*leave*it*on*. Soon the InterneTV could be always-on in the
bedroom, Internet Audio could play in the kitchen, the Internet
News Appliance could display a constant stream of information in
the Home Office, and Cyber-Sears would gladly buy me a transactive
shopping appliance about the size of a catalog. Meanwhile, maybe
they'll finally do some wireless networked thingies right, and
we'll have our own IPCTs for email, vmail, dmail, and real-time
two-way Internet A-V-D.
NECKBREAKERS: Traffic will flood the long-haul lines when the local
access bottleneck breaks. Today, Cable Modems and DSL are the
in-style bottleneck breakers. Behind this, a bar-room of belligerent
technologies is tanking up. At least one of these is virtually
certain to tumble onto the street, broken bottleneck in hand, jagged
and dangerous, pissed at the status quo and ready for mayhem.
Fiber-to-the-home (FTTH) is the clear end-game favorite. It has
been cost effective in new construction for years - pending affordable
terminal equipment and demonstrated, sure-thing uses. Bill's
Billion is riding on Hybrid Fiber-Coax, FTTH's poorer, more available
cousin. Other potential neck-breakers include Data over Electricity,
and a multitude of wireless technologies including LMDS, spread
spectrum e.g., B-CDMA, low-power TV, and Data-From-The-Sky.
Data-From-The-Sky services might be delivered via lighter-than-air
"platform" in the upper atmosphere (Sky Station , Skysat Communications)
or by satellite (ACES, APMT, ASCOM, Celestri, Globalstar, ICO,
Inmarsat, Intelsat, Iridium, Teledesic - did I miss any?). But
the higher the platform, the more expensive the equipment - you
need more powerful transmitters, and/or bigger, more effective
antennas and receivers. So all else equal, the distance to
mass-market equals the height of the platform squared.
GAME CHANGERS? Iridium, Celestri and Teledesic are more than
"platforms." They are entire Networks in the Sky. They
surround the Earth with a meshwork of orbiting, wirelessly
interconnected switches. Their transport facilities traverse no
country. They blur distinctions between local, national and global
service. They make service available everywhere. They
could well change the game.
Iridium will be first (3Q98?), but it is a circuit-oriented,
voice-based network. Then it will be a race to 2002 between two
data-oriented systems: Celestri by Motorola, and Teledesic by
Craig McCaw with Bill Gates and Boeing.
With Teledesic's total global capacity of about
50 Gbit/s (when today one fiber can carry
400 Gbit/s), its service will be premium-priced compared to
terrestrial transport. And end user terminals (telephones?) will
start at $3000 - it is unlikely that Networks in the Sky will get
the worlds three billion poorest people hooked on communications.
But if the device price tracks Moore's Law, and falls by a factor
of 10 by 2002, the richest of the unserved poor could own them.
At least two kinds of positive market feedback could result from
Networks in the Sky. I call them (1) the Hot-Spot Effect, and (2)
the Expensive Chicken Effect. The Hot-Spot Effect would identify
places on earth with lots of satellite activity and attract more
affordable terrestrial facilities to them. The Expensive Chicken
Effect would solve the chicken & egg problem by allowing new,
compelling mobile applications to be discovered - first in
relatively wealthy niches, but spreading to terrestrial
facilities based markets. Both would grow the overall
telecom marketplace.
Networks in the Sky won't be able to be Stupid. They'll have to
be carefully engineered. Over-provisioning will be very, very
expensive. Within those constraints, the simplest networks - with
the fewest assumptions and the most user control - will win.
IN THE LONG RUN . . . Prices will fall, traffic will grow - consumers
and bandwidth providers will benefit. Regulators will churn and
lawyers will litigate, and all that activity will give way to
unchanging human needs, as modulated by the inexorable advance of
technology. New applications will appear, and once every blue-moon
leap-year, between 11:45 and quarter to twelve, one of them will
create new ways to communicate and new markets that take advantage
of them. Let's realize what's temporary and what is inexorable,
and keep a steady hand on the helm.
-------
CONFERENCES ON MY CALENDAR
I will be the luncheon speaker at a NSF-Sponsored Emerging Wireless
Communications Workshop in Washington DC, May 4 & 5. Wireless is
the Stupid Network in action, an end-run on Control-Central. See
http://www.gwu.edu/~cms/wireless/
On May 18, I will be back in Washington DC as lead-off speaker for
a conference called "The Bandwidth Explosion: Understanding New
Technologies That Are Driving Business Opportunities." The website
is http://www.cambridgecom.com/index_conf1.html/
And on May 19, still in Washington DC, I will address the OIDA
Communications Roadmap Workshop. OIDA stands for Optoelectronics
Industry Development Association. The title of my talk will be
"Stupid Optical Networks: A Bright Idea." No conference website.
Find OIDA contact information on http://www.oida.org/
On June 8-11, I will be at Jeff Pulver's Voice-on-the-Net Europe
Conference. I don't know what I will talk about. If it were
tomorrow, I'd talk about capitulating to the government regulationists,
essay above notwithstanding . . . http://www.pulver.com/
On the evening of June 11, in London, I will speak at a World
Technology Network roundtable. A cozy dinner discussion with
technology leaders from the UK. SMART people are welcome.
Contact JPClark@aol.com for more info.
On Thursday, July 9, I have been engaged as Agent Provocateur at
a Washington DC Telestrategies conference entitled "Next Generation
IP Networks." No website yet, but keep checking
http://www.telestrategies.com/
-------
SMART Remark of the Month: "When I read 'Rise of the
Stupid Network,' I was overcome with what I call the moment of DUH!
You know that moment when you come across something that is so blatantly,
painfully, beautifully, obvious and your mind is filled with wonder ....
shortly thereafter followed by 'why didn't I think of that.'"
Martin C. Ahern
-------
Thank you, SMART people, all 506 of you, for being there!
David I
-------
*--------------------isen.com----------------------*
David S. Isenberg isen@isen.com
d/b/a isen.com http://www.isen.com/
18 South Wickom Drive 888-isen-com (anytime)
Westfield NJ 07090 USA 908-875-0772 (direct line)
908-654-0772 (home)
*--------------------isen.com----------------------*
-- Technology Analysis and Strategy --
Rethinking the value of networks
in an era of abundant infrastructure.
*--------------------isen.com----------------------*
Date last modified: 14 Apr 98