Posted: 03/1999

ATM: A Contender for Corporate Networks
By Robert Rosenberg

Today’s networks are in
considerable turmoil, mostly due to market and political considerations rather than
technological ones. However, the result of the turmoil is likely to lead to major changes
in technology as a whole.

Present networks are divided functionally between voice services, video delivery
services, corporate data services and the Internet. Each has evolved differently with its
own set of rules and compromises. Each is moving to take advantage of the newer
technologies as quickly as possible. The lowest common denominator of all of these
networks is at the transmission facility level, where the need is to get large volumes of
bits from one place to another reliably. Whether it is fiber, coax, copper pairs or
satellite, all services require bit bandwidth. The differences are in the services offered
and the way they are offered.

How these networks coalesce is dependent primarily on how the market for the wide
variety of services develops over the next decade. What is apparent is that the demand for
transmission bandwidth will continue to be insatiable, and that no clear winner between
fiber, radio, satellite and the like will emerge soon.

The Disappearing Voice Network

What does all this mean for the circuit-switched voice network infrastructure that has
served us so well for the last several decades? It is rapidly approaching obsolescence.
Within three years, it will be more cost-effective to start converting voice traffic to
asynchronous transfer mode (ATM) rather than to continue to maintain separate networks.

The reason ATM is so much cheaper for voice is that by processing many voice channels
together, much of the control functions that have been focused on 64 kilobits per second
(kbps), or T1, or even T3 channels can be eliminated. At the network endpoints, the
individual channels must be recovered, but in between, the functions of digital
cross-connects and digital switches all will be squeezed into the very high-speed cell
processing of ATM–and, by virtue of very large-scale integration (VLSI)–at dramatically
lower costs. Since ATM has the added capability to send cells only when speech is taking
place, this further will reduce the total voice demand on the network, leading to the
complete elimination of the separate long-haul voice network during the first decade of
the next millennium.

The movement of voice services to new ATM networks flies in the face of the
conventional view that the existing networks will continue to be used and ATM will handle
the growth. An ATM network will be vastly cheaper than an equivalent circuit-switched
network. The newer long-haul carriers with high-capacity fiber optic lines and no old-line
circuit toll switches can handle voice traffic at dramatically less cost than the present
toll carriers who have to recoup their investment in digital voice switches. The major
investment by AT&T Corp., MCI WorldCom Inc. and Sprint Corp. in long-haul voice
switching is depreciating rapidly. It may impact the market value of these carriers
seriously when this depreciation is recognized fully by investors.

The Carrier Perspective

How this sorts out from a carrier’s standpoint has a variety of answers. Some will
continue on the same path, making the best of the present structure. Others will build
parallel networks, while still others will build new networks and gradually migrate away
from the voice network structure. But what structure will they migrate to? There are three
possibilities:

* An Internet protocol (IP)-based full packet-switched network, converting all traffic
(including voice) to Ethernet packets encapsulated in IP to be routed as needed.

* A frame relay-based network, working in a manner very similar to the IP network
except the IP packets are encapsulated to travel the frame relay structure and predefined
routes are established.

* An ATM cell-switching network, segmenting all frame relay and IP packets into cells
and packaging groups of voice calls into cells, all using pre-established routes.

On the surface, the IP-based network approach appears to be the most attractive way to
meet the exploding demand for Internet traffic. It does not require virtual circuits to be
established in advance or have the added overhead of ATM. Further, though IP is plagued by
variable delays at lower speeds as the traffic and speed of the network grows, the
handling of voice gets easier. This is because packet size has an insignificant effect on
delay in very high-speed links, with even the largest packet being transmitted in a few
microseconds (less than 20 at OC-12). However, at these higher speeds, the routing process
becomes a serious deterrent because of limitations on processing power and memory. The
random routing also is still a problem for both voice and video, since the receiving
terminal needs to get packets in the same order as they were sent.

A pure frame relay network approach doesn’t appear likely to become the choice for
carriers because it adds to the complexity of IP routing without helping to solve any of
the other problems, including the effective merger of voice traffic. Its only advantage at
this point is its present strong position in corporate data network connectivity. However,
as Internet demand grows, many corporate networks will move to IP-based intranets, making
IP-based networks more attractive than frame relay.

The ATM network approach easily can handle the explosive growth in traffic, but it
represents a constraint on IP because of the need to establish virtual circuits in advance
and the added overhead of the smaller cell. Nonetheless, it is clearly a superior way to
handle the merging of voice, video and even some IP traffic (such as small packets that
can get trapped behind a bunch of big packets).

The router companies have attempted to counter ATM’s potential to divert carriers away
from pure IP networks by creating new methodologies that do switching of IP packets. Some
of these approaches already have fallen by the wayside because they added too much
complexity to the process while not significantly enhancing the capability. Recently,
rumblings of terabit routers have been heard. If these truly can route IP packets at that
speed, they will be attractive, but it is likely that the cost and/or the complexity will
be unacceptable.

Customer Decision-Making Criteria

Initially, only a small portion of corporate networking will use ATM connectivity
because corporate communications managers are more comfortable with the routed network
approaches such as frame relay. Assuming the pricing for the various alternatives will be
comparable, only those that have chosen to use ATM in their corporate backbone will be
attracted to using ATM service for interlocation connectivity. Later, as frame relay
starts to move to IP networks and as ATM becomes predominant in national and international
networks, more corporate network managers will find it attractive to move to ATM
connectivity. Thus, the promotion of ATM as an effective answer for corporate networking
likely will come from the carriers interested in expanding the use of their core ATM
networks. At that point, pricing for ATM services should be much better than the other
alternatives because it will be adding demand to an already substantial bandwidth base.

Hence, the retail customers’ decision process is not likely to be key to the move to
ATM. Rather, it will be the carriers finding that ATM is the most realistic way to deal
with the exploding traffic demands. If the carriers instead only build IP-based networks,
there will be no place for ATM as a corporate service offering. However, the carriers
likely will choose (and have chosen) ATM for their backbone networks because it is the
only way for them to be insulated from the tremendous expansion in bandwidth demand. Frame
relay doesn’t make sense because its whole structure was targeted originally at T1 speed
access and is having to struggle to reach T3 speeds. Likewise, the IP routing approach is
becoming clumsy as backbone capacity grows to OC-12 and beyond. The leading Internet
backbone providers already have recognized this situation, because most of them use ATM.
Even if a new carrier introduces an all-IP backbone, it will be a small portion of the
total services offering. Even those carriers who do start out with all-IP networks will
find it desirable to introduce ATM in their backbones as demand builds.

The supply of ATM services will come because it is the logical choice for carriers to
handle all the existing services, not because customers find ATM as their only choice.
This is not to say that the carriers’ move to ATM will be easy. An all-ATM backbone will
obsolete a substantial portion of their existing switching and routing investments. In ATM
networks, the functions of toll switching and digital cross-connects are handled readily
as part of the ATM path establishment. Only the local central office (CO) switches will be
necessary to handle existing customer voice services. Fortunately, the transmission
infrastructure will readily handle ATM, frame relay or IP, so most of that will be usable
and easily upgradeable for increased bandwidth.

Robert Rosenberg is president of The Insight Research Corp., a Parsippany,
N.J.-based consulting firm providing comparative market research and competitive analysis
to the telecommunications industry. Visit Insight on the web at www.insight-corp.com.