NNSquad - Network Neutrality Squad
[ NNSquad ] British Telecom says bandwidth costs PRICES unsustainable. True! So why haven't they collapsed?
|
If there is zero money to be
made as a backbone operator why do companies insist on being in that business?
I agree with Stan that this is all a fine mess and even that the reason it’s
a fine mess is that it’s a fine mess. Where I disagree is accepting that
as the only option. All I hear is whining. The problem with the iPlayer
story is that we aren’t being presented with an alternative model that
can deal with the extreme range in the value of bits. But I’ve already
written a lot about how this marketplace is broken and why we should go to an
infrastructure model and how the carriers are threatened by abundance lest we
have a repeat of the fiber bubble. The question is why we are focusing
on not even current but distant past costs as the basis for long term policy,
especially if Enron (or BT) is telling us the price. I’m willing to pay
the bill but I’m not willing to pay the price charged by a monopolist who
hires accountants to support their business models. 10 years is a long time in
Moore’s law – especially when we have so many dimensions for price/performance
improvement. Why should I accept any of
these prices as intrinsic and unchanging when I’ve seen the cost of home
networking go from hundreds of dollars for 2400 bps modems to $50 for a multiport
a gigabit switch in well less than two decades. We’re talking about
million to one cost/benefit ratios or maybe a lot more as we can use better
processing to make more effective use of the bits. If you look at the NVidia
video boards, now being marketed as supercomputer in their own right, we can do
a lot with each bit. Is there any router as
powerful as a video board in a PC that sells for $100 to $500? And why do
routers have to be that powerful – can’t we just use DLP chips
rather than dynamic routing for the backbone of the network? After all, if you
have can assume the route is stable by decomposing the IP address so we separating
routing from naming wouldn’t routing become dramatically simpler? Sure one can argue that a
backhoe operator is expensive but, as noted, we aren’t even using the
current fiber to full capacity. And why lay fibers in isolation – what is
the cost if you’re opportunistic and add a little fiber as part of the
cost of building new roads? And rocket scientists –
that’s like the phone company saying that they can’t afford the
cost of operators. So they put a dial on phones. In the days when I started
doing retail software the culture assumed software was expensive and hard to
support because no one invested in removing all the barriers. Same goes for
home networks. I just don’t buy this whining. Of course I still don’t
understand what is so special about rocket science except that NASA has to be
extremely conservative engineering – the antithesis of the iterative
design that characterizes computing and the Internet. And then we loop back and are
told about the cost of interconnecting with the rest of the network – isn’t
this an extreme form of begging the question? The price is high because the
price is high because … at some point you have to reboot to get out of the
loop. No wonder we want to prevent
China from competing by using patents as a weapon … what if it turns out
that all this stuff can be cheap and simple? Imagine if 19th century
Europe had prevented those Americans from making use of the steam engine and
all those other disruptive ideas? -----Original Message----- Begin forwarded message: From: "Stan Hanks" <stan@colventures.com> Date: September 24, 2008 11:56:54 AM EDT To: <daveb@dslprime.com>, <dave@farber.net> Subject: RE: [IP] British Telecom says bandwidth costs
unsustainable. True? Dave, Saw your note on Farber's list. Unfortunately, Ms. Davis
is right on. *SOMEONE* has to pay for all this... Having built a number of big networks (MFS Datanet,
Genuity 1st generation, Enron) I can absolutely assure you there is
no free lunch. You can cheat a *little* -- a given business case
justifies spending $x, but if you spend a modest amount more you get double
the capacity (not uncommon) but at no point do you get to "all
the bandwidth you can ever imagine using for one low price paid
today". When we build the Enron network in 1998, we laid 2
conduit, and pulled a single cable of 144 fiber in one of them. That meant we
had upgrade capacity of pulling an additional two cables of 144 fiber
in the first conduit plus pulling another three in the second. The
cost for the original build was about $150k/route mile. The cost to
pull each additional cable was about $50k/route mile (unless you
pulled more than one at once, in which case the cost of the next one
went down to about $15k/route mile). OK, let's talk about what that means... of the 144 fiber,
you only use *TWO* per route, one send, one receive, to go into your
DWDM gear. One one route, the Portland-Boise-Las Vegas-Los Angeles
route, we spent over $35mm to light 20 channels of 2.5 Gbs, and would
have spent another $20mm to light the remaining 20 channels
available in that technology footprint. Toni Mack at Forbes did an article around that time
period on the economics of networks (Mack, Toni, 'Empty Pipes,' Forbes,
November 30, 1998, p. 76.). If I recall her numbers correctly, she
pointed out that to acquire an "NFL cities" network, the roughly
25,000 route mile with six strands of fiber (enough for one running system and
four spare fiber for upgrades or emergencies) would cost about
$150mm, but to light it to full capacity would cost over $5B. Yeah.
Billion. Today, 2.5Gbs is passé, and you'd be insane to light less
than a 10Gbs system today. 80Gbs is possible in some areas, but still
not cost effective. And you can get 80 channels typically, so
instead of a 2.5x40 system per pair a decade ago, you can get 10x80
today. And the equipment costs have roughly halved. Moore's Law, if you
will. And we're still just talking about *TRANSPORT*... you
still have to add routers to that in order to build big IP networks.
Alas, the crash in 2000/2001 cost us our best opportunities to do that
cost effectively -- Hyperchip, Pluris, Avici, Procket are all
dead, or effectively dead. Juniper and Cisco are still locked into
a "small box" mentality where you burn over half your
front-side interfaces to build meshes to support your line side interconnects. And
that's before you get to aggregation and distribution networks.
It's just *ugly*... And then you get to manpower, and nothing is cheap there.
And the true rocket scientist guys in building and running big
networks? They're all retired, having either made enough in the boom to go
fishing for the rest of their lives or having just flat burned out... Plus even if you pull that off, you *STILL* haven't
solved the problem of interconnect with the rest of "The Internet"
-- and the rules have changed there dramatically. Instead of "Free
Peering", it's all about parity and advantage. If I don't have something you need,
you typically charge me to take my traffic. And even at the
most aggressive wholesale rates, that's expensive -- rock
bottom market rates are in the $4/Megabit range, and go up from there
depending on volume and how much "pull" you have to offset
the "push"... Back in about 2002 or so, Gordon Cook did a Cook Report
on the economics of big IP backbones, and concluded that even
though no one was actually confirming it, that there was zero money to
be made as a core IP network backbone operator. Having spent the majority of my time since then looking
at "busted companies" and trying to figure which ones are worth
trying to turn around and which should just be broken up, I have to say
that he was pretty much right. The *LAST* thing in the world I'd sign
up for today is being a broadband Internet service provider. There's
just no money in it. Stan -----Original Message----- From: David Farber [mailto:dave@farber.net] Sent: Wednesday, September 24, 2008 12:11 AM To: ip Subject: [IP] British Telecom says bandwidth costs
unsustainable. True? Begin forwarded message: From: Dave Burstein <daveb@dslprime.com> Date: September 23, 2008 9:33:28 PM EDT To: dave@farber.net Subject: British Telecom says bandwidth costs
unsustainable. True? Dave Sally Davis at BT just gave a speech (below) saying BT
might have to block the iPlayer and other Internet video because their
bandwidth costs are unsustainable. Other carriers, like Verizon,
say they have no problem handling the video load, nor expect to have a
problem. So I'm sending her comments over to see if anyone can provide
evidence on why BT's experience is different or that Davis' comments
are unproven hyperbole. It's especially surprising to hear BT's 21st
Century Network, (superbly designed by superb engineers) is
inadequate. Key competitor Sky just pulled off all limits on their $20
broadband service, saying,"it had invested in creating 'a
high-capacity network that is designed to carry huge amounts of traffic without
congestion'" without traffic shaping. I've written BT to doublecheck
the reporter got it right, but the quotes are pretty clear. So am I missing something unique to the UK, or are
Ms. Davis' comments unfounded? Facts welcome. Wholesale giants say Internet will no longer be free ... the answer could be to restrict "free"
access to services like the BBC's iPlayer that allows users to stream BBC TV content
over the Internet ... "One thing keeps me awake at night. In
the immortal words of Jerry McGuire 'show me the money!'," said Sally
Davis, CEO of BT Wholesale. ... Today there are a number of unsustainable
business models out there, and these need to change, Davis
insisted. ... the ISPs are saying "I can't keep increasing the
bandwidth for no more money," Davis said, a situation that will
ultimately lead to ISPs adopting traffic shaping measures and the like to keep
control of bandwidth usage on their networks. ... "We're going to have some very grumpy people,"
namely the content owners and end-users, said Davis. As such, "we have to find new ways around it... Content
distribution models will play a role in that," Davis said.
"We will see those business models emerge," but more work needs to be
done, she cautioned. ... "In the next three years... we will see some
different models emerge," said Davis, a prediction that was greeted
with some scepticism from others in the auditorium. There won't be just one model, "there will be room
for many models," agreed Kathryn Morrissey, EVP at AT&T Wholesale. "Somebody at some point is going to have to pay for
[this network usage]," she said. http://www.totaltele.com/View.aspx?ID=102600&t=2&en=1 ------------------------------------------- Archives: https://www.listbox.com/member/archive/247/=now RSS Feed: https://www.listbox.com/member/archive/rss/247/ Powered by Listbox: http://www.listbox.com ------------------------------------------- Archives: https://www.listbox.com/member/archive/247/=now RSS Feed: https://www.listbox.com/member/archive/rss/247/ Powered by Listbox: http://www.listbox.com |