ATM vs. IP Battle Continues
teledotcom.com
By Peter Lambert and Dawn Bushaus. Peter Lambert is a
senior writer at tele.com. He can be reached over the
Internet atpdlambert@uswest.net. Dawn Bushaus is a
freelance writer based in Chicago. She can be reached
over the Internet atdbushaus@mindspring.com .
No discussion of core network technology would be complete
without the seemingly obligatory-and often misleading-debate
about whether asynchronous transfer mode (ATM) or the
internet protocol (IP) will rule the next-generation public
network. The flames of this debate, which has already lasted
several Internet years, are being fanned again as service
providers develop new classes of service.
When it comes to multiprotocol label switching (MPLS),
ATM would seem to have a head start. Many ATM
switch-makers are implementing MPLS, which may be
because both traditional voice and newer datacentric
backbone networks already have a large installed base of
ATM switches. Further, ATM can deliver standards-based,
guaranteed quality of service (QoS) levels through the setup of
virtual circuits (VCs) programmable for specific performance
metrics, such as variable bit rate (VBR) for bursty data traffic
or constant bit rate (CBR) for delay-sensitive traffic like voice
or video.
These QoS capabilities have remained largely unused in
service provider networks because enterprise customers have
not focused on developing voice over ATM or other
applications to make use of them. Those same enterprise
customers have, however, focused on developing World Wide
Web, voice, desktop videoconferencing, and other IP
applications. Now, it seems, these applications and the
growing demand for differentiated IP classes of service could
finally provide the impetus to implement CBR, VBR, and other
ATM QoS capabilities wherever IP is carried over ATM
circuits.
For these and other reasons, Cameron Sistanizadeh, chief
technical officer for Bell Atlantic Global Networks Inc., says
ATM is the vehicle for starting to build next-generation,
integrated services packet networks this year. Fred Briggs,
chief engineering officer at MCI Communications Corp.,
agrees: "ATM networks are about 12 to 18 months ahead of
routed networks in terms of capability."
The sentiment is echoed by Matthew Bross, chief technology
officer at Willams Network (Tulsa, Okla.), a division of The
Williams Companies Inc. (also based in Tulsa) whose service
provider customers have ATM equipment. "QoS is so
infinitely better defined in the ATM world than in the IP
world," Bross says.
Yet many service providers argue that ATM cell switching will
lose its primary edge-speed-by year's end, as optical
networking and router vendors surpass ATM's top core
speeds by injecting data at up to 10 gigabits per second
directly into optical wavelengths.
Consequently, Qwest Communications International Inc., GTE
Internetworking (Irving, Texas), and others say ATM is useful
to meet some customer demand but unnecessary in the core.
GTE, for one, doesn't plan to use MPLS. "We're looking to
do packet over dedicated links between routers as opposed
to putting traffic into cells," says Dan Wood, systems architect
for GTE Internetworking's Global Network Infrastructure
(GNI). He says GNI will have enough capacity to eliminate
the need for MPLS and ATM for bandwidth management.
GTE's ATM skepticism may explain why its lead vendor,
Northern Telecom Ltd., has taken a 20 percent equity interest
in Avici Systems (Chelmsford, Mass.). Avici's Terabit Switch
Router utilizes ATM QoS algorithms inside its router circuitry
to aggregate differentiated IP service classes, but the company
feeds IP to optics, not to ATM, in the core. "MPLS is
unnecessary and basically promises to sustain sales of Cisco
routers," says Hank Zannini, vice president of business
development for Avici, which has incorporated MPLS into its
software and can "turn it on" if customers require it.
Other providers foresee a mixed environment for the long
term. Bell Atlantic will begin with an ATM-centric core,
Sistanizadeh says, but "there's nothing in the architecture to
prevent us from introducing IP over Sonet or over
wavelengths later, and then ATM is still available for frame
relay and cell relay services." Foreseeing the same migration
path, Williams favors the ATM/optics agnosticism of Argon
Networks Inc. (Littleton, Mass.), whose switch router
Williams is testing, says Bross.
For IP services, that may mean something other than ATM.
"The interesting thing about MPLS is that it wasn't just
designed to fix what's problematic with IP over ATM," notes
Jack Waters, vice president of engineering at Level 3
Communications Inc. (Omaha, Neb.). "As a standalone
technology, it also can put IP on top of any Layer 2
technology. Eventually you could use it to put IP directly on a
wave, so it certainly offers a lot of promise." Indeed, says Joe
Ferguson, director of marketing for terabit router maker
Juniper Networks Inc. (Mountain View, Calif.), "if you can get
40 channels of OC-48 [2.4 Gbit/s] over a single fiber,
somebody would have to explain to me the need for a
congestion prevention tool like ATM."
All use of this service is subject to the Terms and Conditions of Use.
All Rights Reserved.
Copyright c 1998 tele.com
All Rights Reserved.
Website designed by COMPUGRAPHIA
Home page designed by Dennis Ahlgrim.
Last Modified: 8-Jun-98
|
