In For The Long Haul. Profile of Level 3
zdnet.com
Internet 2002 August 31, 1998
Level 3's Crowe foresees 'living' communications
networks based on Net protocols
James Crowe is an old hand at forcing change in the
nation's telecommunications structure. He was founder
and former chairman of MFS Communications Co. Inc.,
a company that Texas Rep. Jack Fields called the poster
child for bringing competition to local phone markets.
MFS was bought in 1996 for $12.6 billion by
WorldCom Inc. A former member of the board of
Qwest Communications International Inc., Crowe left in
December 1997 to start Level 3 Communications Inc.,
which intends to create a nationwide local and
long-distance network based on the Internet
communications protocols. Inter@ctive Week
Editor-in- Chief Tom Steinert-Threlkeld sat down with
Crowe at the company's temporary headquarters in
Omaha, Neb. Crowe explains where Internet Protocol
(IP) networks are headed and why.
What's wrong with today's long-distance networks?
To date, the long-haul networks that I'm familiar with - I
believe this applies to all - have been built like buildings.
There's an assumption about how much capacity you
need to go out and build a network. You take a couple
of years and build it like a project, and at the end you
[think you're done]. Two weeks later, you say, 'My
imagination wasn't big enough. I need a lot more
capacity.'
So, the problem is, networks have been built that
can't be upgraded easily?
Simply put, nobody has designed a network with the
view that it's a living network that has to be continuously
upgraded.
So, how often do you think networks will have to
turn themselves over?
Back at MFS, we had an average asset life assumed at
10.7 years. That was our writeoff. We were very
aggressive. The industry average was like 12 years.
OK. And what will the average be in the future?
Will it be five years going forward?
That's one of the $64,000 questions in our industry.
What's going to be the average asset life, which is
another way of asking, 'What's the equivalent of Moore's
Law in our industry?'
What do you think the equivalent of Moore's Law
will be?
The analog is going to be a reduction in the cost to move
a bit a mile in a second. That will become an exponential.
Moore's Law says that the power of computing
goes up every - doubles every 18 months.
It's an exponential, which is the equivalent of saying that
you'll get about a 33 percent, 34 percent
price-performance improvement every year. That's
Moore's Law.
How many bit miles per second can you buy for a buck
this year vs. next? I don't know if it's 15 percent [less]. I
don't know if it's 30 percent [less]. I know in 1998 it's
going to be a smaller price-performance improvement
than '99. And '99 will be smaller than 2000. And the
reason I say that is while there are components of the
network that are showing exponential improvements, the
vendor community still needs a lot of work to make plug
and play a reality in network components.
You need multiple conduits to put in, theoretically,
an unlimited, replenishable amount of capacity. But
how many conduits will be enough? Are we talking
about half a dozen conduits? Twenty conduits?
The technology to install conduits hasn't been
stress-tested yet. How many conduits? The question is
how many conduits can you get into the ground?
My answer to the question is we want to put in every
incremental conduit we can until the cost starts to spiral.
You don't want to go back and bury [cable] again.
No, and I think those conduits are gold. Particularly if
you think the pace of technological change is going to
increase and not decrease - which seems to me to be an
obvious proposition. And if you think that glass is now
starting to change just as rapidly as the electronics.
All I know is four or five years from now, there will be
some other three- or four-letter acronym that we didn't
see coming that will change the nature of glass, and that
won't be the last. So, it means you've got to have
upgrade paths. It means that you separate the
fast-changing components and transmission and network
gear from the slow. I mean, you put power supplies off
to the side. Put common gear where you expect the pace
of technological change to be slower, and separate that
from stuff that changes fast that you want on a card.
With shortening asset lives, with increasing
competition and fast turnover of technology, what
gives you the confidence then to spend $3 billion
now, and probably $10, $15, $20 billion by the time
you're done with North America - how do you bet
them on IP?
That's a legitimate question. And if you were a
stockholder, you'd say: 'Is this a giant roll of the dice with
3 billion bucks sitting on the table?' And being one of
those stockholders, if I thought that's what it was, I'd sell
my stock and go do something else.
This is no different than the value proposition we had at
MFS: Our fundamental approach is: Make sure your
downside is covered by the value of the asset that you
put into the ground. And the upside can run based on all
of the new services and products you think you might be
able to develop. In the Level 3 case, we think the
downside is protected by the asset value of the national
network, which is going in the ground and is just pure
pipes.
Regardless of what services you provide.
If we just sold the pipes off to people who might be able
to add more value than we can, I don't know if we can
get 90 cents on the dollar back or 2 bucks on the dollar
back. But it won't be zero. So, we think our downside is
locked out by the value of pipes. Plus, there is an implicit
assumption there. I'm sure it is obvious, but it's that
demand for bandwidth is elastic - price elastic. So that
regardless of how rapidly bandwidth is deployed,
demand will go up even faster. Which I believe is proved
in the marketplace.
What is the biggest strategic change that
communications companies must deal with over the
next four years?
In my view, the whole question today is circuit switching
vs. packet switching. And it is apparent, we think, that
you don't have to use generic packet switching, that the
preferred flavor of packet switching is IP. It's all going to
be backwards-compatible with IP version 4. [In
videotape,] VHS is the standard. Well, IP is the standard
for software, hardware and service providers.
What happens if IP version 6 or some other packet
protocol comes along?
Well, [IP version 6] is backwards-compatible. Ten years
from now, we won't recognize IP. We'll have version 6,
and I firmly believe that we will see a merger of some
kind of [data] flow control technology in IP. Whether it's
ATM [Asynchronous Transfer Mode] or one of the
other tag switching [technologies], whatever it is, you are
going to see flow control, which is going to really extend
IP.
Why is that important? Can you not make IP
competitive with traditional telephony without
controlling the flow of data packets as assuredly as
switches control circuits?
Our belief is IP is ready today to fully interconnect with
the public switch network. Act just as you would if you
were a [competitive local carrier] or [a long-distance
carrier], but use an IP network. And that's our strategy.
And what makes you say that it's ready to go
today? I mean, it's ready for data, but arguably not
ready for voice, video or anything time-sensitive.
We think by early next year, by combination of careful
network management and overprovisioning, it's ready for
voice.
Why? How do you overcome the timing-sensitive
problem? Is unlimited bandwidth the answer?
No, bandwidth helps you set up virtual circuits. You can
do it with frame [relay]. You can do it with ATM. And
you segregate your timing-sensitive , from your
nontiming sensitive . It is just a matter of doing it.
Now, the question is, why would you do it? We think it's
because you can get a tremendous cost advantage.
Qwest is offering Qtalk at 7.5 cents per minute,
compared with Sprint [Corp.] or AT&T [Corp.] at a
dime. I mean, I don't find a 25 percent savings on
what is essentially an unknown quality of service
that compelling over a proven service. Do you think
a lot of consumers or businesses will move over for
that kind of savings?
I think history, at least at MFS, would show you that
when you are going up against a competitor who's
entrenched, you can have all the business you want if you
provide a product which is perceived as better in quality
at a 20 percent discount.
So you follow that?
That's our golden marketing rule. Better quality at 20
percent savings. We always started out at a 20 percent
savings, and in plenty of markets we rolled it right up to
parity. After a while, we were in some places the
premium supplier. But you have to start out when you
are the new entrant 20 percent cheaper. Now, why do
we think we can do that? Take a conventional company.
Let's take AT&T, pull out their circuit switch networks,
stick in a packet switch network. Gosh, they can't save
all that much money; therefore, packet switching is not
compelling.
I think that totally misses the point. What that says is,
let's take a mainframe, pull it out. We'll leave all the air
conditioning, leave all the guys in white coats, leave the
card punch. We'll each sit down in job control language,
type up a bunch of punch cards, hand them to the guys in
the white smocks, who will then sit down at PCs and do
the work for us - if that makes sense to you. What I am
saying is, when you get that big a savings centered upon
part of your network, shame on you if you don't get the
benefits that flow through to all the rest of your
operations. I think that you'll fundamentally lower the
cost of your business support systems because,
fundamentally, your network's a lot smarter.
It's smarter because of records you can keep on the
data. Call them packet detail records instead of circuit
detail records. They come off a router today and aren't
even used.
Let's take a good example. Local number portability.
Because nobody trusted Bellcore and the Bells to
develop a local number portability database, [competing
carriers] went out and competitively bid it. Lockheed
[Martin Corp.] got half and Perot [Systems Inc.] got
half. I will guarantee you that was done far cheaper [that
way] than if we'd said like equal access, OK, RBOCs,
go out and develop databases, and let us know what it
costs. Open up those Applications Programming
Interfaces to a lot of smart folks who work on the
weekends and all want to be the next Bill Gates, and
you'll get even more cost reductions. Or apply that same
kind of thinking to all the business support systems and
open network [equipment]. And that's what we are
trying to do. You know a circuit switch is 25 million lines
of proprietary code today.
Where else throughout the network do you get the
good savings? By getting rid of switches?
Obviously, we think that saving is going to start in the
center of the network and push toward the periphery and
eventually, [you're going to ask] what the hell are you
doing with a switch anyway? IP's got an addressing
scheme already built into it. What do you want to
duplicate it with a telephone number for? Basically, when
you run IP over a circuit switch, you've got a protocol
with a nice addressing scheme, and you've got a switch
with a nice addressing scheme, and they're duplicated.
So, push that right to the user. Now, we are not talking
about next year. We are talking a process, not an event.
But what is the point of the circuit switch? All it is is a
way of metering so people can bill by the minute at a high
rate. That is going to come.
How far down will the cost go? Is there anything
numeric you can give us on what we are going to
see, and what kind of margins you can maintain at
lower levels?
I think that we're going into a phase where the costs to
move a bit a mile in a second is going to start to drop at
30 percent, 20 percent, 40 percent, 50 percent a year. I
don't know if it's 20 [percent]. I don't know if it's 50
[percent]. What does that mean? It means the person
who's first to market with the innovative solutions is going
to get a big market share. It means our business is going
to look a lot more like the microprocessor business. And
it means, I think, that you are going to be able to buy a
bit 10 years from now, a bit mile per second, at a tiny,
tiny, tiny fraction of what it costs today. But we'll be
buying huge numbers of them.
How huge? And why?
The way I think about it is, our technological society
spent the last 2,000 years developing systems to extend
our ears around the world. It's only been in the last few
years, at least relatively speaking, that we've been able to
extend our ears at a reasonable cost pretty much
anywhere we want to.
So, now you're going to extend the eye?
Turns out the bandwidth of your auditory nerve is about
a meg and a half [per second] on each side. [The
equivalent of a] T1[phone line] on each side.
What about the eye?
Your optical nerves are interesting. The apparent
resolution in your brain is higher than what the
rod-and-comb structure [of the eye] would indicate -
which means there's some mighty sophisticated software
interpolation going on. But it's in the gigabits per second.
Essentially, what I am saying is: As long as the cost to
move a bit a mile in a second is dropping, demand is
going to go up even faster, until we at least provide the
power of your auditory nerve and your optic nerve to
each user. That's probably a fundamental. I assume we
can't absorb more than what we can get through our
auditory nerve and our optic nerve. But then I am
ignoring machine-to-machine [communication], which is
a whole other source of demand. All of this says to me
that we've got a long, long, long ways for the value, or
the price of a bit mile per second, to drop, and for
demand to increase even faster before we run out of
steam in the industry.
Internet 2002:
Table Of Contents
The Net: Not Just
Data Anymore
Circuit Switching
Vs. Packet
Switching
Laying "Dark
Conduit"
See Me, Hear Me
Qwest For The
Top
Building
Appetites
Beam Me Up,
Craig
O2: Net Lifeline
Highly Charged
Others To Watch
For
Gearing Up
Copyright (c) 1998 Ziff Davis, Inc. All Rights Reserved. Reproduction in whole or in part in any form or
medium without express written permission of Ziff Davis, Inc. is prohibited. Inter@ctive Week and the
Inter@ctive Week logo are trademarks of Ziff Davis, Inc.
|
