The vast majority of the world's telephone and data connections
pass over ATM networks. But does the exuberance over ADSL mean ATM
is about to be put out to grass?
There is a great deal of excitable talk about the next
generation of broadband services. ADSL, Gigabit Ethernet, cable and
satellite modems all promise high data rates for much lower costs
per megabyte.
There are many compelling reasons to purchase these newer
technologies. Most use the existing infrastructure to provide
better performance, from a backbone to a local area network. That
all important backbone technology used by international
telecommunications carriers and big business has also seen new
contenders, such as Nortel's ground breaking OPTera, which provides
between 1600Gbps and 1.6 terabits using an optical D-WDM system.
Yet, despite all this innovation, the most common form of carrier
class communications system is still ATM.With over $3 billion worth
of equipment already installed worldwide, ATM is vital to the
global data communications infrastructure. According to a recent
Dataquest study, the market for ATM backbone products is likely to
grow to $4.7 billion in 2001. Among the biggest spenders in this
sector are companies wanting to connect branch offices to a central
office. Another indirect factor making ATM such a strong technology
is the prevalence of ISDN-B. The symbiotic relationship between
ISDN and ATM stems from the initial installation most telcos
established when data first looked like becoming as important as
voice. Back in the 1980s, the primary data transmission standard
was SDH. SDH offered fast data transmission but with very little
provision for switching between sender(s) and receiver(s). A
harmonised switching technology called cell relay emerged around
this time and together, these were the basis of ATM. For the
telcos, the existing infrastructure was easily adaptable to ATM and
the ISDN services were a great way to divide up these new 155Mbps
and 622Mbps pipes for commercial use. Many business users have some
misconceptions regarding ATM. The biggest is that ATM needs fibre
cabling. This is incorrect. At lower speeds, 155Mbps can easily be
transported over copper. In fact, ATM isn't sensitive to its mode
of physical transport - twisted pair, fibre or coaxial. The
overriding problem with ATM in an office environment is the
comparatively high cost of network cards, hubs and switches. ATM
also has built-in QoS standards, unlike Ethernet, which - up until
the recent 802.3p standard - had very shaky QoS support. Even
today, making an Ethernet network QoS compliant requires a major
overhaul. However, though QoS is built into ATM, it can be much
more complex to implement and configure. Some critics claim that
ATM is not a modern enough standard to support the demands of
multimedia. Chris Hornsey, UK sales director for General Datacomm,
disagrees. " Bringing innovations such as the MAC 500 to the market
place, GDC provides end-to-end provisioning for a wide range of
multimedia services. The MAC 500 (ATM) multimedia access
concentrator distinguishes itself as one of the most forward
thinking and useful products in the voice/video/data convergence
field."The Internet explosion means that people use more video -
and they expect high quality... The product features a modular,
scalable architecture that grows with customers' needs and it
conforms to the latest LAN, WAN and multimedia standards to ensure
interoperability across an entire network." The entry level ATM
market is starting to become more attractive due to the big players
- such as Nortel, Cisco, and Lucent - acquiring smaller companies
while pushing up bandwidth on high-end solutions. The prospect of
OC-768 switching fabric delivery up to 40Gbps can make the current
ATM installation more effective, reducing the cost per megabyte
even further. With these larger dedicated pipes, providers are now
able to offer better bandwidth, both in terms of value and
capacity. ATM does have some problems. When used in a local or
campus network, it is expensive when compared with some of the
high-end Ethernet solutions. However, ATM is quite capable of
transporting IP, so the skill sets required to implement and manage
an ATM solution are not beyond the scope of most IT managers
familiar with Ethernet, Token Ring or Frame Relay. Ethernet offers
highly effective connectivity along the boundary and is simple and
inexpensive. As applications have grown, Ethernet has been scaled
up over the years. From 10Mbps speeds, Ethernet first increased to
100Mbps and then to the modern 1000Mbps Gigabit Ethernet. Many
organisations are providing power users with dedicated Fast
Ethernet connectivity for faster access to server farms and
client/server applications, and, as a result, are relying on the
bandwidth of Gigabit Ethernet to aggregate data-intensive
workgroups into building backbones. The temptation is to spread
this Gigabit Ethernet structure to the whole WAN environment, but
Ethernet has some major problems when compared with ATM. Gigabit
Ethernet lacks ATM's resilience and does not support load sharing.
Without load sharing, even 1000Mbps backbones can become
overwhelmed when supporting multiple Fast Ethernet connections. Nor
does Gigabit Ethernet support true, end-to-end QoS. Although Class
of Service (CoS) traffic prioritisation schemes can be added to the
technology, they increase the cost and complexity of Gigabit
Ethernet and do not provide the performance and latency guarantees
of ATM. As examples, the Institute of Electrical and Electronic
Engineers (IEEE) 802.1p draft standard and the Internet Engineering
Task Force (IETF) Resource Reservation Protocol (RSVP) Internet
draft, when ratified, will provide only prioritisation and
bandwidth reservation respectively. The 802.1p standard will work
only on a packet-by-packet basis at Layer 2; and RSVP, if it ever
comes to fruition, will work only at Layer 3 and will not enforce
the quality of the packet flow. ATM, on the other hand, was
designed from the ground up with the ability to provide full QoS
and traffic management guarantees on end-to-end connections, for
the duration of flows, and without regard to the protocol layer.
ATM has its place within both the enterprise and carrier markets.
Understanding the benefits it brings to tasks which require high
prioritisation or reliability is key. The skill set is not so
complex and ATM isn't sensitive about which transport method it is
used on. Far from being a dead technology, ATM is very much alive
and kicking.
Will Garside