Network Architectures Of The Future
By Carol Wilson, Inter@ctive Week, May 4, 1998
zdnet.com
To that ever-present list of life's certainties, we can now add growth in data
traffic. For, like taxes and death, the demand for bandwidth from data
networks is now inevitable and all-encompassing. So thoroughly will data
dominate the public network of tomorrow that the 20-year plan of the
telecommunications industry to migrate to a single multimedia network may be
shoved aside in favor of network architectures that more efficiently handle
very high volumes of Internet Protocol-based traffic.
Concurrent with the move to data-centric networks is the rapid rise of optical
networking technology. The two trends are not unrelated: Data's insatiable
appetite for bandwidth is driving network service providers and their vendors
to find new ways of increasing available capacity on existing networks.
"There is a convergence of market demand for data, optical networking
technology and the maturity of data technologies for gigabit and terabit
switches all coming into the market at the same time," says Jack Wimmer,
executive director of network technology and planning at MCI
Communications Corp. (www.mci.com) "But it's a juggling act of sorts
because they aren't converging purposefully."
To complicate matters further, as network capacity increases, network
service providers must also find more efficient ways to manage their
high-speed data networks, in order to stay competitive. Thus the push to
manage bandwidth in its purest form -- a color of light, or wavelength -- to
eliminate the clutter of unnecessary equipment and repeated signal
conversions.
"When you are working with 32 channels [wavelengths] of 10 gigabits each,
from an efficiency point of view it makes a lot of sense to do the restoration of
service and the rerouting at the optical layer," and not first convert those
high-speed signals to electrical pulses that can be managed by today's
Synchronous Optical Network (SONET) systems, says Pawan Jaggi,
manager of the optical networking group at Fujitsu Network Communications
(www.fujitsu.com).
The end result of these converging trends is an elegant, optically based,
packet-switched network. But no public network transition on this scale is
ever elegant. Instead, what looms for most network service providers is a
messy and complex process of trying to time technology choices perfectly,
and live with networks that are hopelessly hybrid.
Many Choices
Consulting firm Ryan Hankin Kent looked at possible architectures for a
future network that is dominated by IP services and came up with six
plausible combinations of today's core technologies, including photonics or
optical networking, Asynchronous Transfer Mode (ATM) switching, IP and
SONET. In its report, "The Role of ATM in an IP World," RHK concluded
that choice of architecture depended on the time frame, the applications and
the type of network operator.
"Today's emerging telecommunications world includes a wider variety of
competing and overlapping carriers, a wider variety of vendors and a
tremendous range of technology options," the report stated. "These new
circumstances make it less likely that a single 'winning' platform will prevail."
For each network service provider, then, there will be a series of choices to
be made on how to move today's current networks, many of which are
circuit-switched voice networks, into the realm of packet-switched data
networks and how to shift from today's electrical signals and bandwidth
management to the optical networks of the future. "As operators are making
these choices, we think there are key points they need to keep in mind," says
Carly Fiorina, group president of Lucent Technologies Inc.'s Global Service
Provider business. "We see an issue around what we call network agility, or
making choices that give you options as opposed to constraining options."
It's also clear, she says, that networks must be built that can interoperate with
other networks as standards evolve.
And there is general agreement that different types of services require different
types of networks.
"For networks that only handle IP traffic, using an ATM switch makes little
sense," says Leonard Yanoff, director of advanced systems at Alcatel Data
Networks Inc. (www.adn.alcatel.com). "But if you are talking about a mixture
of services, including voice, then ATM makes more sense for the near term."
Network operators will not be able to avoid making choices, however.
According to MCI's Wimmer, data services will become the driver for
bandwidth by the year 2001 and will continue to ramp up. Feeding into this
trend will be the faster access technologies, such as cable modems and Digital
Subscriber Line systems that are creeping into the market today but may
begin a flood of bandwidth consumption by 1999.
When residential customers, in addition to businesses, can suddenly get
1-megabit-per-second access and more, it will likely trigger a whole new
generation of applications that will capitalize on that bandwidth.
"I don't think we can fully imagine the services that will be there," says
Wimmer.
The very high-speed terabit and gigabit routers now coming onto the market
from companies such as Bay Networks Inc. and Cisco Systems Inc., are
equipped with SONET interfaces that operate at 600 Mbps, 2.5 gigabits per
second or, ultimately, 10 Gbps.
Internet service provider (ISP) GTE Internetworking looked at using ATM
but concluded that the "overhead tax," or 5 bytes per cell used to establish
and maintain the virtual channel end-to-end, was too high a price to pay. "One
of the promises of ATM was that a fixed-length cell makes it easier to support
higher rate interfaces," says spokesman Vaughn Harring. "In reality, the rates
supported by the ATM switches with the set of features required by GTE are
at or behind the rate of interfaces supported by the IP routers -- OC-12c
[600 Mbps] today and OC-48c [2.5 Gbps] very soon."
Different Paths
For operators of existing networks, however, the challenge is very different.
They have to support voice circuit networks and try to migrate a variety of
approaches to data, including X.25 and frame relay networks, onto a single
backbone network to create network efficiencies.
"Cell-based [ATM] networks will be optimum for the next few years because
they do have the quality-of-service features that IP networks lack," says
Derek Underwood, director of systems planning at Siemens Telecom
Networks (www.stn.siemens.com). "We believe ATM fills the window
between today's networks and the network of the future that could well be
based on Internet Protocol."
Nascent voice-over-IP technology might suggest that the move to an all-IP
network would happen sooner rather than later. But most network
technologists are skeptical.
"One of the key things we have to protect is the current services that are on
the network today," says Grahame Rance, general manager for
next-generation networks at Northern Telecom Inc. (www.nortel.com)
"Services such as caller ID, Centrex and other enhanced services don't go
away. Today, it would take a tremendous level of investment to go IP
end-to-end and get the same quality at the same level as 20 years of
development and billions of dollars of investment have done with the voice
network."
What is more likely, says Fiorina at Lucent (www.lucent.com), is a merging of
IP and ATM technologies, as IP begins to take on more of the
quality-of-service characteristics of ATM. At the same time, the
voice-over-IP standards are likely to evolve to include some connections to
the signaling systems that drive today's telephone networks and enable
advanced services, Guess of IXC says.
Optical Push
Just as IP seems a natural dominant networking protocol going forward,
optical networking, or managing bandwidth in its lightwave form, is the natural
direction of physical transmission networks. Already, every major
long-distance company in the country has put Dense Wave Division
Multiplexing (DWDM) equipment into its networks to boost capacity on
fiber-optic cables by adding signals at different wavelengths.
The economics are extraordinarily compelling, Guess says. Using DWDM
and optical amplifiers that can increase multiple frequencies, "we save money
even if we only light up a small percentage of the wavelengths that we add,"
he says, because the company can eliminate multiple regenerators, as well as
add capacity without adding fiber.
Fiorina calls the IXC approach a good example of "getting on the right side of
the technology cost curve," which she says is a fundamental requirement for
today's carriers.
"Carriers need to make the investments today to get their networks positioned
to take advantage of technology changes," she says.
As the amount of bandwidth on fiber-optic cables explodes, today's
SONET-based management systems, which require the optical signal to be
converted to an electrical signal to route traffic, become major points of
inefficiency. Today's optical switching and add-drop multiplexing technology
is still in its infancy, vendors say, but a willing market of network operators is
helping to drive this technology forward at a pace unprecedented within the
public network.
"The need to create very high-speed IP-based networks is going to push the
development of optical network technology," said Denny Bilter, director of
marketing at Ciena Corp. (www.ciena.com).
Few of those building equipment for or operating networks see SONET as
going away, however.
"SONET will be an underlying protocol for optical networks," says Michael
Guess, vice president of engineering at IXC Communications Inc.
MCI's Wimmer sees SONET moving the edge of the network, at the point of
access onto an optical backbone.
That transition could touch off vendor wars, says Joe Bass, vice president and
general manager of lightwave products at Alcatel Network Systems Inc. He
sees manufacturers of customer premises equipment scrambling to add
SONET capability to their equipment even as SONET vendors such as
Alcatel work to add access features to their SONET equipment.
The point of access to the network is likely to be a focal point of technology
development in many respects, Fiorina says.
"We believe it has to be a point of the network where agility is required," she
says. "We think there are going to be a variety of access methodologies --
customer premises equipment is notoriously unpredictable. So, a network
operator can't make a choice that forces you to pick a kind of access over
another."
Who Will Make The Connection?
At the heart of the debate over whether Asynchronous Transfer Mode or
Internet Protocol will be the switching technology of the future is the essential
contrast between connectionless and connection-oriented technologies.
Internet Protocol (IP) is connectionless. Data packets hop through the
network from router to router toward a destination, but there is no path set up
in advance -- the network gives its "best effort" to make connections in a
timely manner.
It's a little like an electronic version of the post office -- when you mail a
letter, it's expected to arrive at its destination, even though you have made no
advance reservation.
Asynchronous Transfer Mode (ATM) is connection-oriented, more like the
telephone network. A virtual path is established when communication is
initiated. It is the ability of ATM to establish these virtual end-to-end
connections that enables the technology to guarantee a quality of service.
These fundamental differences drive the various applications of the
technology.
"You couldn't do what the Internet does today on a connection-oriented
technology," according to Leonard Yanoff, director of advanced systems at
Alcatel Data Networks Inc. (www.adn.alcatel.com).
By contrast, IP-based networks may not be able to deliver the true quality of
service that some delay-sensitive services, such as video and voice, will
require.
"From my perspective, there is always going to be a benefit to
connection-oriented services, and ATM is designed to be
connection-oriented," says Sam Lisle, manager at Fujitsu Network
Communications' (www.fujitsu.com) wideband transport planning
department.
Copyright (c) 1998 Ziff Davis, Inc. All Rights Reserved. Reproduction in whole or in part in any form or
medium without express written permission of Ziff Davis, Inc. is prohibited. Inter@ctive Week and the
Inter@ctive Week logo are trademarks of Ziff Davis, Inc. |
