- Jim
-----Original Message-----
From: Gordon Cook [SMTP:cook@netaxs.com]
Sent: Monday, January 12, 1998 8:18 AM
To: ietf@ns.ietf.org
Subject: Relections on the State of the Net from the Washington IETF and Tokyo
This is the second in what I hope to be an annual series of relections on
the state of the Internet. It is the introduction to my report Building
Internet Infrastructure - published today.
REFLECTIONS FROM WASHINGTON IETF AND TOKYO
In 1997 we saw growth continued, fights over peering, a predicted fiber
shortage that failed to materialize and consolidation taking place not at
the bottom but rather at the top of the market. While prices for bandwidth
from the major providers edged upward, the industry escaped serious price
increases as Qwest and others labored mightily to bring new fiber on line.
With Verio buying up providers and WorldCom going on an acquisitions
binge, consolidation took place but not in the bottom half of the market.
In this context in June AT&T researcher David Isenberg, in his web
published essay "Rise of the Stupid Network." Isenberg articulated with
great clarity and force with respect to the telephone network and the
Internet, what George Gilder has been saying more obscurely for several
years with respect to telecommunications in general. The message: the
Internet is a stupid network with big fast pipes connecting intelligent
devices on users desktops at the periphery. Applying Internet
technologies new companies such as Qwest and new technologies such as wave
division multiplexing are set to create networks offering unprecedented
amounts of bandwidth. In contrast the telephone companies' networks are
intelligent networks designed to ration scarce bandwidth connecting dumb
telephones across a smart network. In this showdown, Isenberg suggests
that the victor will be "the Internet, which, because it makes the details
of network operation irrelevant, is shifting control to the end user."
But the phone companies which are still many times larger than the rapidly
growing Internet are beginning to realize that the Internet has the
potential make their entire infrastructure obsolete.
A critical question is the degree to which the phone companies will
embrace or fight the Internet juggernaught that is bearing down on them.
With the on going wave of telco mergers and a rise of by pass technologies
opening possibilities outside the telco networks, this question is likely
to have a major impact on the Internet business environment in 1998. We
will return to these issues at the end of this introduction.
Peering
In 1995, as the Internet's commercial buildout in the US got underway, the
network engineers were in charge. In a sprit of let's make things work as
widely and as quickly as possible, almost every ISP that connected at the
public exchange points peered with almost every one else. While there
were some technical snafus, they seemed bearable to the engineers.
However, as the largest national service providers all went public, the
territory of the engineers was invaded by the bean counters - those with
the instructions to do what ever was necessary to return a healthy profit.
During the second half of 1996 many of these folk refused to open new
peering sessions with those new comers that did not match their size and
traffic load. In the mean time, about 25 additional national backbones
had been pieced together and the largest five NSPs were each peered with
about 50 other networks. This era of open connections cut more than five
million dollars a year off the cost of what it took to be a national
player by making it possible for those who connected to three or more
exchange points at speeds of 45 megabits per second to build a national
backbone without having to pay the costs of directly connecting to
upstream providers.
Early in 1997 we heard a rumor that turned out to be true. UUNET was
telling its peers that it would not renew their peering agreements and
that they would have to sign non disclosure agreements to find out how
much it would cost to remain connected. We brought the story public in
early May as both UUNET and Sprint announced an end to free peering.
However, by the end of the year relatively few had been depeered. In many
cases UUNET backed off, we were told, because of the bad publicity.
Nevertheless, six months later, John Curran was saying that when up
grades to OC-12 are completed by the five major backbones in late 1998,
he expected the number of fully peered National Service Providers to
shrink from present levels of 35 or so to about ten or less.
For some time the threat of a peering crunch has been worse than the
actual reality. Still the signs are that the rules are going to change and
about two dozen NSPs will have to make serious changes to their business
models if they are to survive. Some will be bought as Genuity was by GTE
Internetworking (BBN) in December. Others will be acquired and disappear
as NetRail did in November.
However, another possibility has recently emerged for these players. This
is the GigaPOP model pioneered by Savvis. Savvis runs what it calls
"private NAPs" in Chicago, New York, St Louis, Atlanta, Dallas, Los
Angeles and Santa Clara. Each of these cities are connected by a DS-3 ATM
Network which serve to channel connections from ten other cities into the
seven private NAP or gigapop cities. In each of these cities Savvis
connects via a DS 3 to MCI , Sprint and UUNET.
This does several interesting things. First it allows Savvis to be a
customer of three of the five largest backbones. Savvis doesn't even
bother to think about peering with MCI Sprint or UUNET. If it did it
would have to do so at the horribly congested public exchanges and would
still need to figure out how to get its traffic to those with whom it did
not peer. Avoiding the public exchanges allows Savvis to give its
downstream customers better performance than those customers could expect
were Savvis to be peered with the big five at the public exchanges.
It also allows Savvis to play the role of a regional aggregator of traffic
including multihoming local ISPs to three of the five major nationals.
It does this more cost effectively than each could do on its own. In
effect Savvis is running seven regional 'airports' for local ISPs. In
doing so it provides a relatively low cost and highly efficient way of
transferring their traffic to the national and international "fights"
(routes) of the Internet.
We believe it to be very likely for those backbones that are threatened
during the coming year by the loss of peering to be able to hunker down to
profitable roles as regional aggregators of ISP traffic for the smaller
number of tier one backbones. We would hope that this regional GigaPOP or
private NAP model will create a counterweight to the power of the national
backbones by aggregating a segment of the ISP market in a way that it
would be more immune to increased costs of connection than it would be if
ISP had to negotiate for individual Sprint, MCI and UUNET connections.
Quality of Service
With the costs of building out major global backbones in the OC3 to OC12
range fall reportedly in the range of hundreds of millions of dollars, the
largest NSPs are under pressure to get business models working that in a
very competitive arena generate acceptable profits. As a result Quality
of Service issues continue to be of great concern. Just as the airlines
would not survive well with only economy class traffic so the commercial
Internet is looking for a business model to provide it with something more
pricey than simple best effort delivery of customers packets - something
that, as Christian Huitema pointed out to us, with current high packet
loss rates is going rather poorly.
In 1996 the industry thought it had the answer in the RSVP protocol. But
1997 saw the failure of RSVP to become a viable delivery mechanism - in
part because it was first thought of three years ago in terms of
servicing video on demand. But, what was needed was something much more
fine grained in order to accommodate the short staccato bursts of web
traffic that have in the meantime become the signature of Internet
traffic.
By the end of 1997 a new answer was emerging: differentiated services
focused on the use of precedence bits in the headers of IP packets. The
integrated services IETF working group mail list created to work in
tandem with RSVP was refashioned along the lines of differentiated
services where there would be "first class" packets that could be charged
for at higher prices.
Service quality issues are taking on an ever broader meaning. In an
article by Gary Anthes in the 12/29/97 issue of Computerworld, BBN's John
Curran is quoted as saying: performance guarantees across different
Internet providers will evolve from metrics developed by industry-specific
groups of users. For example, the Automotive Industry Action Group [see
COOK Report interview on pp. 64 -69 below] is certifying net providers for
adherence to standards in security, reliability and throughput among other
things. Curran says more such industry groups will emerge this year, and
"in two or three years, you'll see a set of specs for business-grade
Internet service that has commonality among them."
IP Telephony Takes a Great Leap Forward
By the end of 1997 voice over the Internet - not video - looked to be the
killer application for differentiated services. It was seen that giving
voice packets preference over the delivery of data would be a cost
effective way of multiplexing Internet telephony applications inside of
customer's regular leased lines. For presumably the only other way to
serve such mission critical applications as voice (a business' telephone
service) would be to purchase separate leased lines just for voice
traffic.
With the introduction of Internet telephony gateways made by VocalTec,
Vienna Systems, and Micom into corporate LANs in 1997, Internet telephony
came of age as a serious business application. Now with the proper
provisioning of a corporate intranet, it was possible to attach the
gateway server to the LAN where it would mediate with the PBX in each
office. The result was that these gateways (costing about $1,000 a port
could take the offices' voice traffic away from the measured use and
billing of the public switched telephone network and place it on the
unmeasured and unbilled corporate intranet. Corporations connected
world-wide via a single Internet provider are beginning to use these
gateways to move all their intracorporate phone traffic off the PSTN and
onto the Internet - something that, especially with international phone
calls, will promise tremendous cost savings.
One can theorize that with almost all the major Internet backbone players
owned by phone companies as the result of consolidation seen in 1997 that
the phone companies may go a little slow in introducing Internet
telephony.
This is not as likely to happen as one might suppose. Why? Because
introducing IP telephony, in many cases, will allow one phone company to
make an end run around the other into markets of its competitors and, once
this happens, the company that has been on the receiving end of the end
run has little choice except to fight back by introducing the technology
as well. Indeed one significant player has been heard to say that the
only way for a phone company to survive the on coming onslaught is to
cannibalize its non Internet based services as quickly as possible.
Unfortunately this revolution may not come cost free. According to Scott
Bradner, in the above cited Computerworld article: In 1998, the transfer
of traffic from the public telephone network to the Internet will invite
the scrutiny of federal regulators who have until now not touched the
Internet, Bradner warns. "It's going to be a wrenching thing when they
start putting regulations, taxes, universal access charges, line fees and
all that on the data service," he says. "I fully expect it to happen
because regulators don't like to see revenue streams go down."
IP Everywhere?
We are hearing for the first time some startling assertions - namely that
such is the momentum behind the IP based interoperable services of the
Internet that it could swallow not only telephony but all of
telecommunications. A few people are beginning to talk about connection
oriented traditional telephone service having only a three to five year
life span.
On December 22 in Tokyo we interviewed Dr. Hiroshi Fujiwara, President and
CEO of the Internet Research Institute. Dr. Fujiwara told us that not
only were Japanese railways and power companies entering the Internet
business, but that many in Japan, and especially the consumer electronics
industry, were interested in a conversion to IP v6 as soon as possible.
Why? Because the consumer electronics industry would like to see every
device it makes have a unique Internet address as soon as possible. This
is something that has long been discussed in a theoretical sense finally
being discussed as a concrete goal.
Bill St Arnaud, the Director of Network Projects at CANARIE, Inc.
(Canadian Advanced Network for Research, Industry and Education) talks of
Internet 2 related research as having heavy emphasis on Quality of Service
applications and with a sly smile mentions what he calls Internet 3 - the
designing of new communication networks from the TCP/IP level downward for
only TCP/IP. He maintains that a part of Qwest's network is being
designed in this way and says that the same holds true for Project Oxygen,
Neil Tagare's attempt to get the major carriers to lay a new swath of
globe circling fiber.
What is beyond dispute is that we are both driving up the amount of data
that a single fiber can carry and laying vast new amounts of the stuff.
At the same time more than half a dozen global wireless systems are soon
going to come on-line. It seems that we have built our way out of any
real shortage and are standing on the cusp of a Gilderian plenitude.
Certainly if all the industry had learned to do was best effort delivery
one might worry about what this new bandwidth would do to industry pricing
in the sense of rendering service so cheap as to cause a pricing collapse.
But, as we have seen, even should bandwidth become as cheap as transistors
are today, we will likely use it in huge quantities. Moreover, money spent
for value added services on top of the bandwidth could come to exceed that
spent for commodity bandwidth itself.
Politics with a
Vengeance
Unfortunately, while our technology brilliance increases, our political
soundness of mind does not. The RBOCs are busily using the travesty of
the 1996 Telecommunications Reform Act to enrich themselves and make the
rest of us pay dearly for having the temerity to think that they would
permit competition at the local exchange level. It is hard to imagine
anything but harm coming from the New Year's Eve decision of Judge Kendall
in US District Court in Witchita Falls, Texas overturning the Act's
Section 271 requirements that the RBOCs show they were allowing
competition at the local level before being allowed to offer long distance
services.
In the meantime the Universal Service portion of the act designating the
availability of funds for public school and library Internet connectivity
is turning into a huge boondoggle. The shame is that Judge Kendall didn't
overturn this $2.25 billion dollar a year tax on rate payers. The schools
and libraries aren't ready. What we have being forced down our throats is
a new bureaucracy headed by a president with a $200,000 a year salary.
And a staff that makes the unconscionable blunder of sending the RBOCs
laughing all the way to the bank by declaring wireless ineligible for the
federal subsidies. Congress or the courts should kill this pork barrel
before it becomes worse.
Unfortunately, this cesspool is not the only political problem dogging the
Internet. Governance or Internet Management as insiders would call it has
become a contentious crisis. The dispute over DNS service created an
IANA/ISOC led attempt at reform which has been sadly mismanaged. On the
surface of things IAHC was a group with a charter that should have worked
to everyone's satisfaction. That it did not speaks to the presumed
arrogance of those who thought that a small committee of Internet 'old
boys" could in a few weeks of private discussion decide late last year
what they wanted to do and then at the last minute go public.
With an ITU member (Robert Shaw), the group decided to involve the ITU in
its process by asking it to be the repository of the IAHC's Memorandum of
Understanding. This decision profoundly unnerved a number of Internet old
hands like Tony Rutkowski who were intimately familiar with the generally
Internet hostile views of the foreign PTTs that make up the ITU's
Governing Council.
The Brian Kahin and Becky Burr led Interagency Task Force on Domain Names
tried to sort out the DNS mess only to almost totally screw up IP
allocation by opposing ARIN. As we watched the bumbling of the task force,
we looked with disgust on the Clinton Administration intervention which,
it seemed, had been created out of fear of the IAHC- ITU DNS solution.
What happened next was ironic because as time passed the IPOC/CORE
reincarnation led by Dave Crocker on the net began to react to its critics
with an arrogance reminiscent of Network Solutions alleged sins as the
sole monopolist for .com. By the end of the summer it seemed that we were
trading one monopolist for another. What's more, Kahin's group of largely
ignorant feds had become convinced that Network Solutions was an evil
monopoly after all. They were now ready to throw the weight of the US
government on the IPOC/CORE scales by legitimizing a highly controversial
effort that claimed the right to register only its self appointed seven
new TLDs.
As we watched Kahin trying to engineer his own solution by picking a group
just as arrogant as Network Solutions could have every hope to have been,
we saw a stampede of registrars to the CORE trough in October. We also saw
other DNS coalitions with working infrastructure being excluded. Finally
we read the Pekka Tarjanne interview where the ITU Secretary General said
that he had concluded after all that the ITU would be the appropriate
governing body for the Internet. We had enough at that point and came out
fully against the IPOC/CORE group.
Fortunately, when he realized how badly Kahin and Burr were doing, Ira
Magaziner entered the frey himself and took over the policy review.
Magaziner has at least had the sense to realize how complex and tough this
situation is and to understand that in October he didn't know enough to
make the right decisions.
number of discussions with the stake holders. He has in our opinion acted
properly to put a road bloc on CORE's forward rush to try to get its names
into the root servers. He is trying to fashion a solution that will let
the Internet remain self-organizing. We are quite impressed at the
reports we get of his level of understanding of the technical details. We
are hopeful that he can pronounce a fair decision and make it stick
through a wind down period running from now until October 1. Policy is
expected to be announced on the White House Web pages before the final
week of January.
Triumph of the Stupid Network
The multi-trillion dollar global telecommunications industry is ultimately
at stake. Companies that didn't exist ten years ago are spending billions
to build a new infrastructure of big, fast pipes to replace the phone
companies' "intelligent" networks. Vast amounts of new fiber are being
laid and rapid advances in wave division multiplexing are increasing the
carrying capacity of this fiber far beyond what was recently thought
possible. Wireless technologies, including spread spectrum, are taking
off. Some seven different "Internet--in-the-sky" systems are under
development with deployment three to five years off. So it seems that
bandwidth - somewhat scarce for a few more months - is likely to become
cheap and plentiful. We see no show stoppers for the present reality --
a fat dumb and happy Internet that connects users who have PCs with the
power of older phone company mainframes on their desk top. The control of
this stupid network is now in user's hands.
Phone companies have two choices: cannibalization of their existing
infrastructure, or to fight and stall and delay to preserve their archaic
empires. In the coming year it will be instructive, with the major
Internet players, now in the hands of phone companies, to see which
companies try to lead and which try to fight the changes. We argue that
this communications revolution will be won by those who offer users the
best set of cost-effective dumb pipes. We also argue that with companies
like Qwest determined to become Internet providers themselves, the
pressure on the phone companies not to stick together in resisting change
will be great.
The choice is likely to be one of joining the revolution or being swept
away by it. To survive, countless executives will have to overturn their
own world views. We argue that success will come to those who adopt new
technology and use it quickly to cannibalize existing operations. These
people must decentralize their control over the choice of intelligent
applications by placing them into hands of end users who then will drive
the pace of change by running them over a stupid network.
January 12, 1988
You have just read the introduction to
BUILDING INTERNET INFRASTRUCTURE
Evaluating Technological, Operational, and
Architectural Drivers of Growth as Players
Consolidate & Bandwidth is Set to Explode
Volume Three of an Annual Handbook on the Commercial
Internet's Business, Technology and Management Issues
An Anthology of Recent Articles from The COOK Report
For complete information for this 312 page report, published today, please
see www.cookreport.com/building.html
************************************************************************
The COOK Report on Internet New Special Report: Building
431 Greenway Ave, Ewing, NJ 08618 USA Internet Infrastructure ($395)
(609) 882-2572 (phone & fax) www.cookreport.com/building.html
cook@cookreport.com Index to 6 years of COOK Report,
glossary, executive summaries, special reports found at www.cookreport.com
************************************************************************