4/98 tele.com article. Internext: This Revolution will Be Televised
[Good article on plans of many ASND customers. One specific ASND reference related to UUNet and UUCast.]
teledotcom.com
By Peter Lamber, Senior Writer
OC-3. OC-12. OC-48. OC-192. OC-192 with WDM. And
on and on. Bandwidth multiplicity is the obvious answer to
keeping the Internet one step--or maybe it's a half-step--from
being overtaken by the volume of its success. At least it's the
answer that most service providers are relying on now to stave
off disaster. Yet even as they plot out exponential growth
plans for Internet capacity, leading network builders realize
that someday they are going to lose the race--unless they do
something to change how the race is run.
"It's getting to the point where I need not to just believe in
miracles but to rely on them." That's what Vint Cerf,
acknowledged sire of the Internet and senior vice president of
Internet architecture and engineering at MCI Communications
Corp., told fellow Internet builders at a technology conference
in February. Cerf's prognosis: Without those miracles the
Internet could be in for a major brownout by year's end. The
scary part of Cerf's prediction is that today's traffic volumes
are almost nothing compared with what will literally come
down the pipe in the next few years.
The kind of video content now broadcast over the air and
over cable TV infrastructures will dominate the world's
telecommunications networks by 2005. Content providers like
The Walt Disney Co. are moving ahead with R&D that will
enable them to deliver any number of video services to any
number of subscribers over the public network. Their ultimate
goal is to obliterate today's barriers--such as the geographical
limits of over-the-air broadcasting and the economic limits of
dealing with cable providers--that separate them from a truly
global mass market.
That begs one enormous question: How can network
infrastructures deliver infinite content if they can't even keep
pace with the relatively puny amounts of traffic coursing
through the Internet today? Which leads to another question:
Are folks like Vint Cerf really putting their fate in the hands of
divine intervention? Not quite. Even as they continue to
hammer more capacity into their backbones and send silent
prayers to the cybergods, Internet providers are taking real
steps to redefine exactly what the Internet is, how it works,
and what it will deliver to the masses. Over the next few years,
these efforts will manifest themselves into something that may
be as unimaginable today as the World Wide Web was a
decade ago: the Internext, a network resource that could
make today's Internet data networking's equivalent of two tin
cans and a string.
Some of the technology and network design components that
will enable the Internext already are off the drawing board and
ready for deployment in the real world (see "Internext Building
Blocks"). Effectively drawing a blueprint of these Internext
components, new entrants into the Internet business, including
the eventually IP-based cable industry, have begun to plot out
a new public network architecture from scratch. In terms of
fundamental physical design, that blueprint begins with the
decentralization of Internet content, using intelligent database
replication tools to push copies and updates outward to data
center server farms and closer to users in every locality. The
plans also include implementation of mature but long-dormant
IP multicast routing protocols, laying the groundwork this year
and next for a network that can broadly support not only the
one-to-one unicast transmissions that drive almost all 'Net
traffic today but also one-to-millions multicast transmissions.
Redesign is being carried out beyond the backbone as well.
Local service providers are deploying cable modem and digital
subscriber line (DSL) access technologies that are not only
high speed but also "always on," pointing the way to new
categories of real-time, anytime services. To control traffic
patterns and congestion, leading internetworking vendors are
starting to deliver a tangle of competing, currently proprietary
quality of service (QoS) routing protocols that, in two to five
years, could make the 'Net intelligent enough to accommodate
each unique class of application with a unique level of
transmission quality assurance.
Finally, the grand design for the Internext points to a
communion of intelligent signaling, directory services, and
object-based application management technologies that
promise to link users, service providers, and content more
flexibly and seamlessly than ever.
The first manifestations of the Internext already have started to
appear. For example:
* Cable Internet providers At Home Corp. (@Home,
Redwood City, Calif.), US West Media Group (Englewood,
Colo.), and Time Warner Cable (Stamford, Conn.) are
overbuilding dial-up IP service networks with networks that
deliver more bandwidth at the edge than in the core, replicate
content in dozens of local markets, and try to solve Internet
congestion and performance limits by turning today's
bandwidth-to-storage spending ratios on their heads.
* UUNet Technologies Inc. (Fairfax, Va.) is overbuilding its
own unicast Internet backbone with a discrete multicast
dial-up backbone to support one-to-millions services.
* GTE Corp. is building a national IP network employing QoS
routing intelligence that GTE believes will provide performance
good enough to overbuild not only the Internet but also the
public switched telephone network in one swipe.
* US West Inc. has made deals with the biggest names in
computer application, database, and directory services
software both to spawn "network-resident" applications and to
extend the public switched network's intelligent signaling and
service provisioning systems into the Internet.
Internext pioneers are sweating the details now not just to
survive short-term but also to create completely new business
models and revenue streams. A multicast IP infrastructure
would allow Internet service providers to multiply the breadth
of entertainment and information channels, just as cable TV
multiplied broadcast TV's breadth, creating a true mass
market for interactive video services at last. A distributed
infrastructure that supports a chain of global, national, and
local product distribution partners could replace the current
kings of commerce with new wholesale and retail royalty--a
vision in which ISPs take over the clearinghouse role, but with
the advantage of delivering any number of services to any
number of customers. An infrastructure intelligent enough to
allocate its resources one user and application at a time would
enable ISPs to steal premium customers not only from media
and commerce giants but also from the 150-year-old public
switched telephone network as well. A directory-enabled,
object software-enabled infrastructure will create an entirely
new role for the ISP as application storage house for business
communities of interest to share. Indeed, after years of
industry preoccupation with technology, the business interest
horse may finally be moving in front of the cart.
The trouble today
Today's grim reality for Internet backbone builders is that even
the exponential bandwidth increases promised by
wavelength-division multiplexing (WDM), gigabit or terabit
Ethernet, and other technologies won't keep today's 'Net from
melting down. Cerf projects that the number of Web hosts
may more than triple this year alone, to some 100 million
machines, and that the number of Internet users may reach
300 million by midyear, up from 70 million in January.
Conventional PC-based Internet usage is just the tip of a huge
iceberg: International Data Corp. (Framingham, Mass.)
projects that 89 million non-PC devices, including TV
set-tops, screen phones, and mobile assistants, will be
connected to the Web by 2001. Throw bandwidth-hungry
multimedia applications together with user, device, and host
growth, and the pressure on the legacy Internet infrastructure
multiplies geometrically.
More bandwidth alone will soon become insufficient to keep
service providers afloat. MCI doubled its backbone capacity
last year, but Cerf told attendees at February's IP Multicast
Summit in San Jose, Calif., that bigger backbones are no
magic bullet. "I'm getting nervous about the router side of it,"
he said.
In fact, it's the changing nature of Internet traffic that has
planners spinning into white-knuckle mode. "First it was simple
ASCII files, then Netscape brought HTML graphical files, and
then it was streaming audio and now video, each step adding
pressure to bandwidth and servers," says Robert Bowman,
director of backbone engineering for Exodus Communications
Inc. (Santa Clara, Calif.), a provider of distributed content
colocation and Web hosting services. "Even providers of
e-mail are coming out with feature after feature, like
multimedia attachments. As an end-user, my experience is
getting better and better. As a network engineer, I see the
pressure on the infrastructure."
Neither faster switches and routers nor WDM are enough to
meet projected increases in backbone capacity demands, says
Milo Medin, chief technology officer for @Home. The
answer? "Get away from dumb pipes and build a smart
network," Medin says.
Ultimately, that smart network will do much more than cope
with Internet traffic as we now know it. It's also going to
reshape the role of data networking in general, a reshaping that
will alter the very business models on which service providers
are now based, says Martin Hall, chief technology officer for
Stardust Technologies Inc. (Campbell, Calif.). Stardust
manages the IP Multicast Initiative (IPMI), a consortium of
vendors and ISPs supporting the implementation of IP
multicasting transmission standards.
First-generation ISPs sold basic connectivity, Hall explains,
while second-generation providers are distinguishing
themselves with services like Web hosting and IP faxing. "The
third generation is about a business perspective shift to selling
content," Hall says. "ISPs are realizing they're in the content
distribution business, so they need to look at broadcast, cable,
and satellite distribution models, leveraging subscription and ad
revenues and eventually electronic commerce."
If that sounds a lot like the old rationale for interactive TV,
that's no accident. "The Internet has actually made interactive
TV viable," says Scott Watson, chief computer scientist for
Walt Disney Co. Research and Development (Burbank,
Calif.), which serves ABC TV, The Disney Channel, and other
divisions of the media giant. "Internet broadcast will do to
cable what cable did to broadcast, except this time it's 5,000
channels. It's important to begin arguing through the business
models and their infrastructure implications now."
The fun starts
Those arguments are in fact heating up among service
providers and their suppliers. "We are clearly moving beyond
the world of point-to-point to a world of point-to-multipoint
communications," said Ed Kozel, outgoing chief technology
officer at Cisco Systems Inc. (San Jose, Calif.), to IPMI
members in February. "The endgame is multimedia to millions
of people. This is the market we're targeting through our
technology developments and all our partnerships."
To get there, Kozel insists, infrastructure providers will have to
change the ways they move information. Last year, new
entrants and specialist ISPs began to implement a half-dozen
such fundamental changes, such as decentralization of content
storage, experimentation with service classification and tiered
service routing mechanisms, deployment of multicast router
and access device software, and creation of multimegabit
residential connections and always-on residential Internet
connections.
One of the most fundamental changes now under way is the
decentralization of Internet content storage. Leading this
charge are the cable ISPs, which aren't burdened by legacy
internetworking gear. Since 1996, @Home, US West's
MediaOne Express, and Time Warner's Road Runner have
been building (and, rumor has it, could soon combine) their
own Internet backbones linking dozens of regional data
centers. Servers in each data center house copies of regularly
updated popular content made accessible to local cable
modem users.
Over the past year, the same local storage model has come to
rule network designs among ISPs specializing in content
colocation and hosting, including Exodus Communications,
GTE Internetworking (Irving, Texas), Globalcenter Inc.
(Sunnyvale, Calif.), and Digex Inc. (Beltsville, Md.). Three of
these ISPs have been acquired by telcos in recent
months--GTE Internetworking used to be BBN Planet Corp.,
Globalcenter was bought by Frontier Corp. (Rochester,
N.Y.), and Digex is now owned by Intermedia
Communications Inc. (Tampa, Fla.). Last month, Sprint Corp.
unveiled its own distributed IP Web hosting service, further
confirming industry consensus that these ISPs are moving in
the right direction.
The new distributed networks correct a key shortcoming of
the legacy Internet by placing content closer to users. That
distribution cuts delay, improves other performance factors,
and minimizes use of precious backbone capacity by allowing
millions of users to travel only down the street, rather than long
distance, to visit Web sites. "Bandwidth is cheaper and
performance is better when you make it local," says Bowman
of Exodus Communications. "A London user coming all the
way to California for a Web page is silly."
Medin of @Home says the distributed architecture breaks
with old Internet thinking. "It uses a lot more processing in
storage, rather than more bandwidth," he says. Translation:
Internet providers can rely on the steadier, proven technology
advances in computing power than on less predictable
exponential increases in network bandwidth. "We decided to
follow the computer curve rather than the telecommunications
curve," he says.
The result for @Home is a network design that's diametrically
opposed to the Internet. "@Home actually has the most
bandwidth at the edge, less regionally, and even less in the
backbone," Medin says. "The network itself must decide
which data is available locally, and we can do that because of
intelligence at every layer."
The Internext's intelligence extends beyond distributed content
to include increasingly intelligent protocols for discriminating
among classes of applications and allocating network
resources on a session-by-session basis. Cisco, Bay
Networks Inc. (Santa Clara, Calif.), and other network
equipment makers have begun to offer IP QoS mechanisms in
their router software. Those mechanisms, which operate both
within and above the network layer (Layer 3) of the Open
System Interconnection (OSI) model, include the ability to
define high, middle, and low classes of service and to "color"
or "tag" each IP packet with a priority classification. Routers
then recognize a packet by its priority color bit, effectively
recognizing the application in the packet for the purposes of
special treatment.
A cast of millions
The endgame for all this activity is to figure out a way to
virtually limitless quantities of data--including interactive video
services--to the mass market. There's absolutely no question
that today's Internet is a technological dead end--witness the
performance problems caused by last year's killer application,
push technology. Compared with the envisioned interactive
video services, push amounts to a single drop in the demand
bucket.
The very real threat of traffic overload is prodding Internet
builders to revisit and deploy the IP multicasting protocols
developed a decade ago by the Internet Engineering Task
Force. IP multicasting utilizes UDP/IP, instead of the dominant
TCP/IP unicast protocol, to transmit a single file or stream to a
list of network destinations. Guided by a complete list of
subscribers' IP destinations, the single stream leaves a server,
then splits itself repeatedly wherever a router table confirms
that down this or that tributary lies at least one recipient on the
list. Current implementations can reach several hundred
thousand addresses with a single transmission, and vendors
promise to reach millions this year--a process a million times
more bandwidth-efficient than delivering a separate unicast
from one server to each of a million users.
The MCI Real Broadcast Network venture with streaming
media software provider RealNetworks Inc. (Seattle) uses
multicasting to distribute streaming media to local servers for
on-demand, rather than live, access. "In the next year or so,
there will be a huge explosion of 100- to 500-kbit/s content,"
says Martin Dunsmuir, general manager of software for
RealNetworks. Only multicast networks can scale to handle
such traffic, he says.
One such network is UUCast, announced late last year by UUNet Technologies. UUCast uses multicast-enabled Cisco
routers and access concentrators from Ascend
Communications Inc. (Alameda, Calif.) to deliver a single
multicast stream to up to 300,000 dial-up Internet access
ports. America Online Inc. (Dulles, Va.), AudioNet Inc.
(Dallas), and Microsoft Network (Redmond, Wash.) are
among early UUCast users, primarily for entertainment and
distance training applications. Multicasting also is making its
way into corporate networks, via technologies from
companies like StarBurst Communications Corp. (Concord,
Mass.), Starlight Networks Inc. (Mountain View, Calif.), and
Precept Software Inc. (Palo Alto, Calif.), which is being
acquired by Cisco.
Applications--whether they're entertainment for the masses or
multimedia corporate offerings--ultimately will become the
stake driven through the heart of the Internet as we know it.
Hammering that stake will be low-cost, mass-market,
appliances--slim clients designed not to imprison applications
in boxes outside the network, but to borrow them from inside
the Internet itself.
Tele-Communications Inc. (TCI, Englewood, Colo.) may
have delivered the first stroke in December, when it
announced deals with General Instrument Corp. (Hatboro,
Pa.), Sony Electronics Inc. (San Jose, Calif.), Microsoft
Corp., and Sun Microsystems Inc. (Mountain View, Calif.) to
take shipment of 5 million to 11 million OpenCable digital
TV/cable modem set-top boxes over the next five years. The
OpenCable architecture is multicast-friendly, since cable
modems provide "always on" connections.
"Always on" is destined to be the mantra for content providers
determined to blow away today's Internet service business
models. "We've been trained by TV to get entertainment,
leave it on, and incur no usage tariff," says Disney's Watson.
"The appliance attached to the TV and the notions of the
network computer and device-independent applications are
fundamental to the new IP services vision."
With huge companies like Sony and Microsoft sharing TCI's
market-creation risk, much attention is being paid to the
potential of OpenCable's open architecture to draw mass
development of applications for blending digital TV and IP
content. But the larger implications of "thin," low-cost,
consumer-friendly Internet appliances like the OpenCable
set-top and Microsoft's WebTV may be not only new
consumer services but also a radical increase in how much
content and application management responsibilities ISPs must
embrace.
"In the OpenCable world, everybody who watches digital TV
gets an e-mail address by the end of 1999 or early 2000,"
says Adam Grosser, product development vice president for
@Home. "This is about ubiquitous information into the home,
which might translate to an Internet-connected kitchen recipe
pad. You create a different set of motivations for potential
customers." Grosser believes that such a model also reduces
consumer inertia toward getting connected by no longer
forcing consumers to understand how to install PC software,
respond to application error messages, or manage network
connections.
And there's the rub for ISPs: The network must take over
those application management jobs from the client. That means
having a protective layer of application execution middleware
residing in the customer box. Rather than physically installing
applications via disk or CD-ROM at the box, the consumer
will rent an application--say, a parental control content filter or
tax return program--from the network itself. The application is
downloaded from servers to network appliance memory for as
long as it's needed and is executed by a middleware virtual
engine, such as Sun's Java Virtual Machine.
In this brave new world, the ISP must now manage not just
content, but also applications, stored and managed by the
network. Internext pioneers are now getting ready for this day.
"The distribution chain is the absolute core of @Home,"
Grosser says. "It's about creating new distributors, replacing
Blockbuster video stores with video on demand or Barnes &
Noble with amazon.com. We've built the tracking and billing
infrastructure to do that."
Translation: The Internext isn't just on the horizon. Pieces of it
are already here.
All use of this service is subject to the Terms and Conditions of Use.
All Rights Reserved.
Copyright c 1998 tele.com, The McGraw-Hill
Companies, Inc.
All Rights Reserved.
Website designed by COMPUGRAPHIA
Home page designed by Dennis Ahlgrim.
Last Modified: 10-Apr-98 |
