Core Switch Makers Ponder An IP Future
August 17, 1998
Inter@ctive Week via NewsEdge Corporation : The new
service providers that aren't saddled to legacy
equipment like to talk about Internet Protocol as the
future of networking. But, at least for now, they're
making sure to deploy Asynchronous Transfer Mode
technology to offer services.
That's good news for the companies that make
Asynchronous Transfer Mode (ATM) core switches.
But even as these manufacturers rack up sales of their
ATM products, they are starting to develop long-term
plans to incorporate more Internet Protocol (IP)
technology into their product lines.
Level 3 Communications Inc., Qwest Communications
International Inc. and Williams Communications Inc. are
among the new service providers that have tested and
installed ATM switches as part of the core of their
fiber-based communications networks.
"As a wholesale provider, we need to take into account
the needs of carriers that have all types of media in their
networks," says Amy Reiber, senior manager at Williams
(www.wilcom.com), which uses GX550 ATM switches
from Ascend Communications Inc. (www.ascend.com) in
its network. "It's important to use ATM because it offers
quality of service," she says. "There's also the ease of
use it gives customers who are connecting back into our
network."
In many ways, these new carriers have no choice. ATM
is a proven technology that can handle different forms of
traffic effectively without delay. Its ability to offer
different qualities of service enables providers to make
service guarantees and offer differentiated services with
differentiated pricing -- something that IP can't do just
yet.
ATM's advantage on that front won't last forever,
however. "It's only a matter of time until IP is perfected,"
says Niall Gallagher, senior manager at Northern
Telecom Inc. (www.nortel.com), a provider of ATM core
switches. "As IP networks get faster, the issues of jitter
and delay are going away."
As the performance gap between IP and ATM shrinks,
the debate over which technology to use will become
more intense. The big issue is whether investments made
now in ATM will look like the smart move five or 10
years from now.
"We've reached a funny stage," Gallagher says. "To
compete in offering IP services, providers have to build
ATM services."
The Big Switch
The coming together of the ATM and IP worlds has
equipment makers scrambling to articulate a long-term
strategy. Most manufacturers say they expect the
network's core to feature a mix of ATM and IP --
although some aren't ready to say exactly what that mix
will be or how it will work.
"We're looking at IP very seriously right now, says Joon
Miller, director of product planning at Fujitsu Network
Communications. "We're looking to develop an IP hybrid
product. We're not sure whether we'll be IP- or
ATM-centric, but it has to be something in the middle,
like an IP router and ATM switch, to address that
market."
To date, however, Fujitsu (www.fnc.fujitsu.com) has yet
to announce an IP strategy for its core products.
Ascend continues to preach ATM. "Some vendors
would love you to believe that everything is going to be
IP," says Tim Krasky, vice president of marketing for the
core systems division at Ascend. "Well, SNA [System
Network Architecture] and frame relay and voice are not
IP. ATM is the best bet for those multiple media types."
The key advantages of today's ATM core switches are
size and predictability. ATM core switches are big and
typically based on distributed processors with
high-speed interfaces and direct fiber connections.
They're built to carry large volumes of data at incredibly
fast speeds. And with voice migrating into the core, the
core itself becomes larger.
These monster switches do need some intelligence,
according to Krasky, such as being able to reroute traffic
automatically around a failed route through the network
via Switched Virtual Circuits. "There's no market for
stupid switches," Krasky says.
Some Of This, A Little Of That
Several ATM core switch makers say they will develop
terabit IP switch routers to accommodate IP traffic in
their networks. But those products won't necessarily
replace ATM at the network core, says Fred Baker,
chairman of the Internet Engineering Task Force
(www.ietf.org).
"It's a mix of engineering trade-offs," Baker says. "If you
need a muxed interface that can talk to everything, then
that'll be your choice. If you need a point-to-point
relationship between two routers, then packet over
SONET [Synchronous Optical Network] is OK. They'll
both have a place in the sun."
Baker notes that today's terabit products aren't ready for
heavy-duty core action. The segmentation and
reassembly chips that handle conversion between IP and
ATM formats aren't available at OC-48, or
2.4-gigabit-per-second speeds, he explains. "Until we get
a chip to handle framing at OC-48, we can't even talk
about terabit speeds," he says.
One knock against ATM is its high overhead when
compared with IP. With ATM, five bytes of every
53-byte cell is reserved for header information.
"Service providers say eliminating the cell tax could pay
for the line, " Baker says. "There are costs and benefits
to each approach."
As the performance gap between IP and ATM shrinks,
the debate over which technology to use will become
more intense
Core Curriculum
Here's how key Asynchronous Transfer Mode (ATM)
core switch makers plan to handle Internet Protocol (IP)
Vendor
ATM Core Switch
IP Plans
Ascend Communications Inc. (www.ascend.com)
GX550
Will develop terabit IP switch router
Cisco Systems Inc. (www.cisco.com)
BPX
Sells TGX terabit switch router
Fore Systems Inc. (www.fore.com)
ASX4000
Will develop hybrid ATM/IP core switch
Fujitsu Network Comm. (www.fnc.fujitsu.com)
FETEX-150
Will develop terabit IP switch router
NEC America Inc. (www.nec.com)
ATOMNETM20
No IP plans announced
Newbridge Networks Corp. (www.newbridge.com)
MainStreetXpress 36190
Will develop hybrid ATM/IP core switch
Siemens AG (www.siemens.com)
MainStreetXpress 36190
Will develop terabit IP switch router
<<Inter@ctive Week -- 08-14-98>>
[Copyright 1998, Ziff Wire]
|
