Internet2: Do we really need it? [ASND R&D reference]
Excerpt: "UUnet and other big data networking companies are
putting bucks behind innovative start-ups, such as Juniper
Networks Inc., which is developing routers capable of
handling a trillion bits of data per second - a new threshold.
And they are plowing big bucks into development. Just five
companies - Ascend Communications Inc., Bay
Networks Inc., Cisco, Fore Systems Inc. and Newbridge
Networks Corp. - are spending $1.8 billion this year on
research."
msnbc.com
Builders of the current Net question the focus, utility of high-profile government-funded projects
By Randy Barrett, Inter@ctive Week Online, ZDNN
May 4 - Internet 2: Either you love it or you hate
it. It's the latest vision for a high-speed network
connecting academic and research institutions.
But don't confuse it with the Internet itself. The
builders of the now commercially funded and
developed Internet aren't so sure it's even
necessary.
Gore debuts shape of Net to come
Internet2 and the Next Generation Internet are two completely
different initiatives
University Corporation for Advanced Internet Development
Next Generation Internet Initiative
THE CLINTON ADMINISTRATION GAVE
Abilene - the new high-speed network proposed by the
University Corporation for Advanced Internet Development
- a high-profile send-off at the White House on April 14.
Amid flashbulbs, grins and handshakes, government,
industry and academic leaders promised to deliver vast
technical improvements to the commercial Net.
"This project will help develop technologies for the
Internet that are faster and more reliable," said Vice
President Al Gore, the man who made "information
superhighway" a household term. "They will be instrumental
in the creation of the next-generation Internet."
But out in the real world, commercial Internet engineers
range from skeptical to downright scornful of government
and academia's big promises to better the Internet. Both the
Next Generation Internet (NGI) initiative, federally funded
to the tune of $100 million this year, and Internet2, funded
by a consortium of companies and universities under
UCAID, elicit far more suspicion than praise in the
commercial Internet community.
"Internet2 is of no value to us," said Alan Taffel, vice
president of marketing at UUnet Technologies Inc., the
largest provider of Internet backbone services. "[Its]
funding will end up supporting a private network with little
benefit to the Net at large."
UUnet and other Internet backbone provision
companies insist they can already provide private and
secure networks for almost any use. Already, UUnet's
backbone operates at 622 million bits per second - equal
to the speed of the current federally funded network
connecting supercomputing centers. Qwest
Communications International Inc., which has agreed to
supply $500 million worth of communications capacity to
the Internet2 project, is building a cross-country fiber
network that can ship data at the rate of 10 billion bps.
"We're coming to
a time when we
don't need a
government-funded
backbone."
- RICK WILDER
Senior manager for Internet
technology at MCI
Communications
"I see no benefit at all. I see it as a distraction," said
Robert Laughlin, president of DataXChange Network Inc.,
who said he wonders why the government needs to get
involved in advanced Internet Protocol (IP) development in
the first place.
WHAT'S TO SHOW?
Many commercial network engineers said they just
haven't seen any tangible benefits recently from federally
funded Internet research - including grants to universities
using the Very high-speed Backbone Network Service
(vBNS), an academic network started in 1995 by the
National Science Foundation.
"It's not really bringing us anything new. We've
reached a point where the commercial Internet is beyond
the research and engineering backbone," said Rick Wilder,
senior manager for Internet technology at MCI
Communications Corp.
Wilder has firsthand knowledge: MCI operates the
622-million-bps vBNS for the NSF. "We're coming to a
time when we don't need a government-funded backbone,"
he said. MCI already offers 622-Mbps service to its
commercial customers.
Internet2 and NGI supporters concede there have
been fewer transfers of groundbreaking new technology to
the commercial sector in the recent past.
"There hasn't been as much dramatic stuff happening in
the last five years," said Richard Mandelbaum, chairman of
NyserNet Inc., a regional nonprofit backbone.
The "golden era" of tech transfer among government,
academia and the nascent commercial Internet was between
1986 and 1994, when 20 years of accumulated research
and development into advanced networking, packet
switching and the Internet protocols burst into more
mainstream usage and attention. During the same time
frame, the Domain Name System arrived, as did HyperText
Markup Language, the Mosaic browser and the
all-important Border Gateway Protocols for network
routers.
"On the private
side, we have the
best money
funding the best
minds, and that
ultimately has got
to advance the
Internet more than
anything in the
public sector."
- ALAN TAFFEL
Vice president of marketing
at UUnet Technologies
These historic transfers of technology from the research
community to the commercial Net are widely lauded by
production network operators, but much of the appreciation
ends there.
PRIVATE SECTOR RULES
"The government should declare success and go solve
health care," said UUnet's Taffel, who is convinced the next
burst of key technologies will come from the private sector.
"On the private side, we have the best money funding the
best minds, and that ultimately has got to advance the
Internet more than anything in the public sector."
Commercial demand, for instance, has forced
equipment vendors in the past three years to condense by a
factor of 100 the number of ports that can be squeezed into
a single rack on an Internet access device at an Internet
service provider's network point of presence. Cisco
Systems Inc. and UUnet also worked, for instance, to
develop a router that could effectively handle multicasting,
the technique whereby a single audio or video file is sent
through the Net, and copies are pulled off by machines that
have subscribed to the broadcast. And commercial
companies have pushed the rate of transfer for frame relay
packets to 622 Mbps, from 45 Mbps.
UUnet and other big data networking companies are
putting bucks behind innovative start-ups, such as Juniper
Networks Inc., which is developing routers capable of
handling a trillion bits of data per second - a new threshold.
And they are plowing big bucks into development. Just five
companies - Ascend Communications Inc., Bay
Networks Inc., Cisco, Fore Systems Inc. and Newbridge
Networks Corp. - are spending $1.8 billion this year on
research.
But it isn't as if academicians don't understand the
need for speed.
"It's very irritating. It assumes the educational research
folks [academicians] have no idea what's going on in the
real world," said Steve Coya, executive director of the
Internet Engineering Task Force, the body that develops
and creates new technical networking standards. "It's a
very myopic point of view."
Scott Bradner, senior technical consultant at Harvard
University, agreed: "They forget where they came from.
NSFNet [the original research network started in 1986]
was the proof of concept for high-speed networking. There
is a real need for what Abilene and NGI are bringing to the
table."
"It's a very
myopic point of
view. They forget
where they came
from. NSFNet
was the proof of
concept for
high-speed
networking. There
is a real need for
what Abilene and
NGI are bringing
to the table."
- SCOTT BRADNER
Senior technical consultant
at Harvard University
Bradner and others see Internet2 as a test bed for
cutting-edge multimedia applications that require not just
speed, but coordination over long distances. These are
applications such as industrial-strength digital libraries,
virtual "collaboratories," where scientists around the world
can work together on complex experiments, immersive
environments that push the bounds of reality, and online
studies of the origin of the universe.
George Strawn, director of advanced networking at the
NSF, admitted there have been few world-shaking transfers
of technology out of federally funded research since 1995,
when the agency started granting high-performance
connection awards to universities using the vBNS: "My
view is that, by and large, we're still in the research and
development phase," he said.
A look at the period from 1995 to 1998 shows a
handful of less glamorous protocols found their way from
the research community into the commercial Net.
Determining their exact paternity is difficult, since some
stemmed from federally funded work, while others arrived
from collaborations between vendors and schools. The list
includes Squid, a widely used protocol for caching
information on networks; RSVP, for reserving bandwidth
for special uses; and a code called Weighted Fair Queuing,
which lets routers handle packets more efficiently.
Researchers said they expect more to come from the
vBNS, since the initially strict acceptable-use policy was
modified 18 months ago to allow more schools to use its
bandwidth.
THE FUTURE NET
Engineers of the commercial Internet said their primary
concern is scaling up huge networks in the face of enormous
demand and troublesome congestion. They aren't interested
in new high-bandwidth applications for distance learning,
multimedia and advanced research collaboration - all key
goals for the Internet2 program.
Douglas Van Houweling, president and chief executive
officer of UCAID, said they should be: "That set of issues
they're grappling with is why Internet2 came into existence,"
he said.
Internet2 was launched by a consortia of U.S.
universities in 1996. UCAID was created in October 1997
to oversee the research and funding of Internet2 activities,
including creation of the Abilene network, which will run at
9.6 billion bps when fully operational in 1999. Currently,
120 schools and 25 affiliate organizations are members of
UCAID.
Van Houweling and other Internet2 advocates said that
only by developing and testing high-bandwidth applications
on a research network can they learn how to move today's
Internet beyond its bandwidth and scaling challenges.
"(Fundamental
change) won't
come about from
incremental
change in the
commercial
Internet. They're
looking too closely
at the floor."
- GEORGE SMIT
Director of network marketing
at Northern Telecom
That means new router protocols that can direct huge
volumes of bits per second accurately and correctly. In the
future, a few dropped packets could mean the loss of
billions of bits of data. Both routers and switches need to be
redesigned to handle the load.
"[Internet2] will be an experimental network like the
Internet was 15 years ago," said David Lytel, president of
NyserNet, a nonprofit network that serves the university
community in New York state. "We need an Internet that
can break."
The existing vBNS connects 46 universities. Some
schools connect their own advanced networks with the
vBNS through high-speed connection points called
gigapops. UCAID plans for the new Abilene backbone -
named after a historic Texas railhead - to connect up to
30 gigapops together to form a new network that allows
higher speeds and more bandwidth. Abilene will also
connect to the vBNS; UCAID will get the new bandwidth
free from Qwest.
Internet2 research promises to include more of the
private sector than the predominantly university and
governmental contributions that led to Internet 1. UCAID
has in its membership 25 leading infrastructure vendors.
All technology created under the UCAID banner is
open and available to the membership, and the corporation
will not charge royalties. Annual corporate dues cost
$10,000 per company.
Internet vendors that are UCAID members insist the
Internet2 effort is necessary and worthwhile. Stephen Wolff,
executive director of advanced Internet initiatives at Cisco,
said he expects key breakthroughs from the group,
particularly in the area of enhanced quality of service.
"If you point to the past, a lot of really neat things have
come out of this community," said Wolff, who also directed
Internet research at the NSF in the early 1990s, when the
Net became commercially viable.
ISPS UNDER THE GUN
Wolff is convinced that big advances in networking
won't come from the Internet service providers. "That's not
what commercial operators are about," he said. And, while
they might even bump up against the Abilene network in
terms of speed, the character of that traffic is vastly
different.
George Smit, director of network marketing at
Northern Telecom Inc., said his company has a special
interest in Internet2 research into high-speed access as well
as quality-of-service improvements. Smit said fundamental
advances are needed in the way the Internet is managed,
particularly in differentiated services.
"It won't come about from incremental change in the
commercial Internet," he said. "They're looking too closely
at the floor."
Given the impressive track record of university and
federal research on the creation of Internet 1, some
onlookers said UCAID and NGI officials have done a poor
job of justifying and promoting their research programs to
engineers in the trenches.
"We don't think the two universes are that far apart,"
Smit said.
|
