IP: August COOK Report on Internet Published

[Dave Farber <farber () cis upenn edu>]

From: Gordon Cook <cook () cookreport com>




LEVEL 3 BUILDING 16,000 MILE FIBER NETWORK TO LINK URBAN BUSINESSES AS
DID MFS UPGRADABLE STATE OF THE ART NETWORK TO BRING MOORE'S LAW TO
IMPACT TELECOMMUNICATIONS COST, pp. 1- 5


We interviewed Jack Waters, Level 3 Vice President of Engineering, about
Level 3's plans which, as was the case with MFS, include linking urban
businesses with high speed TCP/IP inter city transport. Unlike MFS,
Level 3 will become its own inter-city carrier via a 16,000 route mile
national fiber network. It will use Peter Kewit & Sons to build the
network along the rights of way of Burlington Northern, and Santa Fe
Railway as well as, the Union Pacific.


Level 3, which is leasing fiber from Frontier, while it builds out its
own fiber, is a year and a half behind Qwest in the construction that it
is about to begin. Unlike Qwest, it will directly connect large urban
buildings to its inter-city network. It will lay at least 6 conduits to
make its network easily upgradable.


While Waters agreed that Level 3's LEAF fiber could run without SONET,
he points out that SONET will be there, for the most part, because of
the mission critical nature of Level 3's voice traffic. (In talking to
others we have ascertained that SONET will also be installed
ubiquitously, when it otherwise would not be needed, because DWDM
technology pushes so much bandwidth through a single strand, that SONET
is needed to allow carriers to provision pipes of sizes smaller than OC-
192. For example, OC-12 and OC-3 represent bandwidth that could not be
provisioned without SONET muxing and demuxing equipment.)


Level 3 will be using its leased fiber to begin offering voice and IP
data traffic by the fall. Its plan is to make Its network steadily
upgradable in order to take advantage of the latest capabilities and
cheaper prices driven by Moore's Law. Level 3 will be heavily involved
in IP-based telephony that it will be able to offer for a fraction of
circuit switch prices.


The company intends to create an infrastructure that first and foremost
interconnects with the current PSTN at a peer level. According to
Waters: when one thinks about peering with the PSTN, one must be
thinking about connecting to the PSTN at the edge - - much in the same
way that a CLEC would interconnect to the PSTN today. "We think that it
is very important not to change user behavior when creating this new
infrastructure and offering our services over it," he says. "Therefore
Level 3 is not talking about dialing around the PSTN or anything like
you may see today with voice services using Internet style technology.
It is talking about connecting directly to the PSTN and offering voice
over IP service not dependent on anything like auto dialer's or pin
codes that would involve a significant change in user behavior."


In this context, Level 3's most important early contribution to the
market place may be the TAC effort launched on June 22. The Technical
Advisory Committee composed of eleven premier hardware and software
companies plus Level 3, was formed to develop a set of technical
standards to bridge current circuit based PSTN and emerging IP networks.
The protocol is expected to be published, as well as submitted to global
standards bodies including ITU, ETSI and IETF within the next two weeks.
The TAC is hoping for rapid adoption by the standards bodies. Member
companies intend to begin shipping products based on the IPDC protocol
by year's end. Readers may expect more about this in the September COOK
Report.


PEERING CONTINUES TO PERPLEX -- NO NEW PEERING AMONG PRIVATELY
INTERCONNECTED BACKBONES IN MORE THAN YEAR - ADJUDICATION OF COMPLEX
ISSUES TEMPTING BUT UNLIKELY TO PRODUCE POSITIVE RESULTS FOR THE
INDUSTRY        pp. 6 - 13




Peering continues to be a messy situation. Undoubtedly, the big five, as
early entrants to the Internet game, have advantages even over extremely
well-financed upstarts such as Level 3 and Qwest. Furthermore, given the
sorry state of public exchanges, the only way for new players to get
adequate connectivity and performance is to either be privately peered
or become customers at multiple points. We believe that the worst
features of the current situation are twofold: the secrecy in which
peering is shrouded by non disclosure agreements and the fact that there
are no generally accepted rules. These two features can combine in
insidious ways to cause newcomers to the market to fear that they are
being cheated. We hope that some way to alleviate these problems can be
found. As the discussion that follows shows, while the position of the
big five is defensible, the murkiness of the world in which it exists
creates a temptation for those who are locked out to pursue legal
action. While we still sympathize with the smaller players, we are leery
of legal action being able to produce any result other than moves that
will lead to the regulation of the Internet.


We republish a lengthy discussion from the Cybertelecom law list.
Requiring that peering agreements be published is seen as one way to
blunt the anti-competitive aspects of the current situation. This
suggestion is made that peering arrangements are part of the terms and
conditions under which backbone providers are offering
TELECOMMUNICATIONS services as defined in the Communications Act.
According to a list member: "they undertake to ship data of the user's
choosing (packets from the user's own customers) to the places the user
wants them to go (the addresses indicated on the packets) with no net
change of form, content, etc. If this is true then peering arrangements
are subject at least to regulation and probably to actual tariffing."
After much discussion, the consensus was that peering really doesn't fit
in such a Procrustean bed.


DYNAMICAL OPTICAL SWITCHING NOT YET ON HORIZON FOR OPTICAL NETWORKS
pp. 13, 24


In a Letter to the Editor discussion, Mike Trest and Bill St Arnaud both
agree that the new optical backbones will be fast and dumb.


PROBLEMS IN EXCHANGE OF INTERNET TRAFFIC AND NETWORK PERFORMANCE ISSUES
DEBATED -- HOW TO KEEP LOCAL TRAFFIC LOCAL? IS METERING AN ANSWER?
pp. 14- 16


Discussion of benefits of local interconnection and problems of
assessing distance sensitive charging or any other kinds of metering.


WHO SETS OPERATIONAL ROUTING POLICY?    pp. 16, 20


Some complaints from Australia about the obscure ways in which global
routing policy is set.


BEHIND CLOSED DOORS & WITH IANA BACKING USPS FLOATS PLAN TO TAKE OVER
.US TLD --KAHIN AND IANA SUPPORT ATTEMPT TO OFFER ALL CITIZENS
ELECTRONIC MAILBOX IN CONTEXT OF CONTINUED NON DISCLOSURE BY BURR'S TASK
FORCE AND ITAG          pp. 17-20


We publish the US Postal Service's Confidential Memo detailing why it
wants control of the .us domain. We complain that USPS and IANA should
be discussing this policy in public. We debunk false Internet.news
(MecklerMedia) report of alleged discussion of USPS plans at Reston GIAW
meeting. We conclude with cautiously optimistic meeting assessment. This
meeting has become the first of a series to be held during the next six
weeks to attempt to establish consensus for the by laws and make-up of
the Board of the new IANA corporation. It now looks much less likely
that ISOC and Jon Postel's ITAC can impose its own design on the
Internet as a fate accompli.


SEAN DORAN SUPPORTS RED AS INTERNET'S MOST IMPORTANT BANDWIDTH
MANAGEMENT TOOL p. 20


Sean Doran explains his efforts to implement RED and shows the benefits
that result.


WORLDCOM'S PROPOSED DIVESTITURE OF MCI BACKBONE FOUND WANTING IN LIST
DISCUSSION      p. 21


An extremely perceptive discussion in which Sean Donelan states: MCI is
selling C&W the most expensive portion of their customer base.
ISP/wholesale customers tend to be the least loyal, and have the highest
line utilization and most complex routing configurations. Which in a
fixed-rate environment translates into the highest cost/revenue = lowest
profit ratio. Since this is the reverse of how most telco regulators
think, I don't know how this will go over. In the telco world, high
usage customers are good things because more minutes of use means more
dollars of revenue. But in the Internet world ISP customers are a good
way to boost your backbone usage, but they also boost your backbone
expenses ALOT. On the other hand, corporate customers (which MCI isn't
selling) tend to be the most profitable, sometimes only using a part of
their circuits from 8-5 during the day. And most corporate customers
tend to be reluctant to change providers.


SPRINT'S INTEGRATED ON DEMAND NETWORK ASSESSED  p. 22


Discussion on the NANOG list finds the concept interesting but notes
that details of implementation are unlikely to become clear before next
year.
***************************************************************************
The COOK Report on Internet            New Special Report: Building Internet
431 Greenway Ave, Ewing, NJ 08618 USA  Infrastructure ($395) available. See
(609) 882-2572 (phone & fax)           http://www.cookreport.com/building.html
cook () cookreport com                    Index to 6 years of COOK Report, how to
subscribe, exec summaries, special reports, gloss at http://www.cookreport.com
***************************************************************************