ATM and IP Fight for supremacy
totaltele.com
ATM, for so long touted as saviour of the legacy telcos, is
slugging it out with upstart IP for control of the world's
telecomms networks. Paul Gannon watches the battle unfold as
the two standards struggle for global domination
Only a few short years ago telecomms equipment
suppliers and operators were touting asynchronous
transfer mode (ATM) technology as the answer to
everything. However, a direct comparison can be
made between the ATM experience and what
happened to an earlier technology: ISDN.
Like ATM, ISDN was hyped as the only technology that would allow operators
and users to adopt a single communications protocol to carry all manner of
communications - voice, data and image - over digital links.
However, as history shows, ISDN failed to live up to all the hype and, in fact, is
now little more than yet another niche technology, - widely adopted by some
countries and widely ignored by others. In essence, the carriage of multiple
communications streams turned out to be far less important than ISDN's ability
to offer digital transmission speeds in excess of most modem technologies.
Basically, ISDN failed to realise its potential because other cheaper and/or faster
technologies matured more rapidly in the corporate computer world - despite the
fact that ISDN links can be bundled in multiples of the technologies' basic 64
kbit/s capacity.
Disregarding the lessons of history, ATM apologists were sure that ATM
experience would be so very different to what happened with ISDN. The
rationale? - that ATM is scalable in speed and that telcos have learned about
marketing; that they now now understand customer needs; the pricing of digital
services, and the prime importance of the provision of applications to run over a
transport technology.
That was then, yet even now (and with the benefit of hindsight) many telcos and telecomms equipment suppliers are still reciting the old mantra. "ATM will see me into retirement," said one 30 year-old marketing expert from a major European telco, "there's nothing else on the horizon to challenge it."
Well, that's as maybe, but outside the telecomms world the ATM message is much less clear. "What is the point of ATM?" mused a representative of an
Internet service provider (ISP) - answering himself by adding, "The telcos have
spent a lot of money on ATM and want to foist it upon us. Whereas what we
really want is IP connectivity."
When worlds collide
Over the last five years the Internet Protocol (IP) has risen to challenge ATM
head-on. Until very recently suppliers were assuring us that ATM would become
the dominant protocol in local area networks - however, the IP lobby believes
that the siren song of ATM is now less compulsive; ironically enough, as a direct
result of the customer power to which suppliers pay such frequent lip-service.
The pro-IP people believe their favourite technology will push further and further
into the telcos core wide area networks and reduce ATM to the ranks as an
insignificant also-ran.
As usual, the truth probably lies somewhere between the two extremes The core networks of the telcos are likely to be centred on ATM while corporate networks will be dominated by IP. It is the point of intersection - where the two worlds collide - that is pivotal to the debate, and the question must be whether ATM can reach beyond the edge of core networks and into the customer premises equipment (CPE) or if IP will penetrate the core network, leaving ATM only on the busiest links.
Among telecomms equipment suppliers and telcos alike there is widespread agreement that the main strength of ATM is its ability to offer different classes of service to different transmissions. Indeed, it is this ability that enables it to carry different types of traffic so efficiently.
For example, data transmission is less susceptible to delay than voice and image, but is usually 'bursty' in nature. Thus, data requires high bandwidth for a very short period and ATM was specifically designed to handle this sort of traffic.
IP, meanwhile, has its origins in data transmission and is less than ideally suited to voice and image transmission. The packet-based technology offers variable cell size, which, in comparison with fixed-length cell-based technologies such as ATM, introduces inefficiencies and requires extra processing power.
Furthermore, IP is currently a single quality of service (QoS) protocol and while it usually doesn't much matter if e-mail is delayed by a few seconds, it is imperative that real-time data traffic, as well as voice and video, goes through immediately.
Currently, voice-over IP (VoIP) has two modes - passable quality over private networks with substantial bandwidth, or inferior quality service via ISPs and public telco networks.
Internet standards bodies are upgrading IP to be able to handle different classes of service and ensure that higher quality voice, video and real-time data transmissions are possible. However, "if you put all these ingredients into IP then you are really making ATM version 2," says Markku Hellstrom, of Tellabs in Finland. "And what is the point of re-inventing ATM when it already exists?"
Testing times
The answer may lie in the different approaches taken by the telecomms and computer worlds. In telecomms the emphasis has always been on quality assurance and new services have traditionally been regarded as less important than guaranteed up-time.
By contrast, in the computer world, companies rush products to market and often get users to buy 'beta' versions - thus ensuring that customers themselves do the testing. This is because computing is built around the premise that technology which becomes 'the standard', is not necessarily the best but the one with the biggest market share.
In telecomms, standards emerge differently, usually as the result of the slow but co-operative convergence of notions about how to implement a technology - which is only later released onto the market. The net effect is that, whilst telecomms standards are limited in scope, they also have a history of being extremely effective.
And this is the arena in which the IP and ATM battle is joined. Today, whilst the the telcos are arguing that ATM is a good standard that should be implemented, the computer industry is steadily building a critical mass of IP installations that could eventually overwhelm the telco's position.
"IP comes via the Gigabit Ethernet router world and simply cannot provide the same QoS that telco grade switches can," said Stewart Davies, head of network product design at British Telecom. "Phone outage tolerance level is zero - and this is something the computer world will have to learn. Cisco and similar companies come from an industry where technology is not designed for non-stop running - and telecomms customers won't accept questionable levels of quality of service.
Davies adds, "We are not unaware of IP - and, furthermore, we're not the laggards that people suggest. The real issue is the application of old telco thinking
to the new world of business. We do have to learn to think like new network
operators, but there are pitfalls. For example, the BT Internet service as built on
the computer industry principle of getting a product to market as quickly as
possible,however, it collapsed, and we then had to rebuild it with telco-level
quality of service."
Indeed, BT is now migrating its various separate data networks - frame relay (FR), leased variable bandwidth, switched multi-megabit data service (SMDS) - onto a single 'multiple service platform' which will carry all of them over ATM.
When this is done, BT will offer IP services over the ATM infrastructure. In such a case VoIP will simply be a replacement for leased line business and therefore not particularly important.
For the ATM Forum, ATM is at its most effective when it is transparent. "Customers may never perceive that that they are buying ATM because it is hidden at the service level," says David Wills, chair of the Forum's European market awareness committee. "They may think they are using FR, but then most FR and IP traffic in the US is carried over ATM backbones."
Dave Gellerman, vice president of carrier marketing at Newbridge, points out that telcos have been slow to provide the services that corporations and slower still in providing them at competitive prices. The result? Users have built private networks to carry voice, then data traffic.
The telcos can, indeed do, push ATM as a transport for corporate IP traffic. However, delays in the offering of value-added IP service will allow other players into the management of corporate VPNs - and telcos could find themselves relegated to a role as providers of commodity bandwidth.
"Up to 75% of European companies will have an intranet by the end of this year," says Dave Gellerman. "The question is whether they will build private networks."
Mighty fibre
Some new network operators are proposing to build fibre-optic networks and offer large corporations direct fibre connections linked to high-speed routers. In
the early stages such operators will be able to offer only limited coverage and
their customers are more likely to be ISPs than users.
However, as fibre laying continues and as wave division multiplexing (WDM)
technology permits fibre capacity to be increased by factors of tens, hundreds
and perhaps even thousands, it should soon be feasible for corporations to lease
dark fibre or use a managed service providing direct end-to-end IP service. The recent alliance between Cisco and leading WDM supplier, Ciena, indicates that we may be about to see that market tested.
For telcos to regain the initiative they will have to offer end-to-end IP over ATM. Some, such as Telia of Sweden and Telecom Finland, have pioneered this approach in Europe. Elsewhere, Telecom Italia has released its MondoNet IP VPN service to carry multiple IP services to users. And, in the US, legacy infrastructure-free new carriers, including Qwest and Level 3,will offer pure IP
transmission services.
On the supply-side, companies such as Nortel and Nokia are turning their attentions to IP-switching. Peter Boland, vice-president of Nortel's Webtone subsidiary, says that things are changing so quickly that telcos could miss the switch to IP.
"Six months ago I would have been a firm ATM advocate, but IP is growing so fast that it will soon have enough volume to make ATM unnecessary," he says, adding, "Were I to build a network today I would still use ATM, but in two years' time networks definitely will be IP dominated."
Boland believes that carriers should prepare to migrate as the need arises. He dismisses the concern that building QoS parameters into IP to support different types of traffic is no more than re-inventing the ATM wheel.
"Yes, the Giga-routers of the future will look like ATM switches because they will tackle the same kinds of issue, but if we designed ATM today it would be a lot simpler," he said.
In his estimation, the present traffic mix is 95% ATM and 5% IP. However, the ratio will be 50/50 in two or three years and to 10/90 in favour of IP in five years. "If the telcos don't prepare for this they are going to get killed" says Boland.
ATM heavies
Andy Bray, marketing manager at General Datacomm Europe and representative
of the ATM Forum has a different point of view. He says, "The notion of a
confrontation between IP and ATM doesn't hold - it's no longer a matter of
choice. The point is that IP needs ATM technologically since it is a solution to
the scalability problems of IP but ATM is commercially beholden to IP, because
IP has such dominance in the marketplace. ATM's future could be as an
integrating technology without any native applications at all."
Others remain insistent that ATM will dominate. Indeed, some big ATM
developments are currently underway. For example, BT plans to deploy 200 ATM switches over the next couple of years to support its multi-service platform and both France Telecom and Deutsche Telekom have also announced similar large-scale exercises. [ASND business??]
In Germany, the research firm, DFN, has 120 ATM switches from General
Datacomm in a network providing IP and other services. PTT Telecom of the Netherlands also has a network with over 100 ATM switches. Hutchison Telecom in Hong Kong has a large ATM network, as do the telcos in Malaysia, South Korea and Singapore.
To date, one of the most successful new entrants has been MCI Worldcom (through its MFS and UUnet acquisitions). It is firmly wedded to the use of ATM as a transport mechanism and has gone so far as to suggest that IP services will be made available through UUnet rather than WorldCom itself.
Protocols developed within the ATM Forum, such as LAN emulation (LANE) and multi-protocol over ATM (MPOA), are effectively restricted to LAN interconnection over limited numbers of nodes. However, the Forum maintains that although MPOA originated as a campus standard, it is also applicable to wide area networks.
Suppliers such as Ericsson with General Datacomm and Bay Networks, and Siemens with 3Com and Newbridge, are also working on a wide area solution: multi-protocol label switching (MPLS). The standard is being developed by the Internet Engineering Task Force (IETF).
MPLS combines layer 2 performance with layer 3 connectivity and network services, promising to reduce the complexity and cost of other IP/ATM connectivity options. It also enables the QoS options essential to multiple transmission types. "MPLS is attractive to carriers as it will start making ISPs more interested in ATM," said General Datacomm's Bray.
Newbridge's Gellerman says: "We will use MPLS signalling to get information to the user over ATM. The telcos also have to offer VPN IP capability to suck IP traffic into the carrier network. One way to do this is to use digital subscriber line (DSL) technology to deliver ATM as a transport for IP to the CPE."
Newbridge recently released its own line of ADSL products designed to extend VPN services, transported over ATM, via high-speed digital lines directly into CPE. "It is an ideal way of delivering a remote managed port," said Gellerman. "Providers can use ADSL, running at either 6Mbit/s and 640kbit/s, to offer
multi-tiered revenue streams, charging say US$50 a month basic ADSL rental, in
addition to a VPN and other service fees."
A further bonus for carriers is that such links can be managed as part of the core ATM network, offering customers flexible bandwidth and services.
Gellerman predicts that, in the longer term, corporate solutions will not be sufficiently adaptable to provide the economies of scale that will be possible on carrier networks.
However, others take a different line. Thus, a Tellabs spokesperson says: "There is no single right way to do things, it depends on the network. If there are a lot of
legacy services, ATM is better for migration, but in a buildout from scratch other
technologies may be better."
The fight for supremacy in the telecomms industry is as much about positioning,
share and future market control as it is about standards. To date, the predictions
that new entrants would spot and exploit opportunities faster than the incumbent
telcos have proved correct.
Ultimately, the matter will be settled by the endemic differences between the
computer and telecomms industries. Computing has always taken the greater
risks to capture market share and when IP reaches critical mass, as it soon will,
ATM could find itself in the same technological cul de sac as ISDN.
Going underground
Passengers on the Berlin Underground can
now use the new 'emergency pillars' which
have been built on the metro platforms. A
press of one of the buttons on the pillar
instantly connects the user to help and
information personnel, while the exchange
is monitored by surveillance cameras.
The new system is the largest yet
ATM-based video surveillance system and is part of BVG's (Berliner
Verkehrsbetriebe) move to improve safety on the U-bahn. All 168 stations on
the nine lines are networked, as are the 230 video cameras. "We call it wide
area private network," says G,nther Bartsch, product manager, broadband
communications at Siemens, which is provisioning the project.
Since the topology of an underground railway rules out a mesh configuration,
the network has been specially designed. Forming the backbone ring is 160km
of fibre-optic cable. The ATM switch nodes at each station use add/drop
multiplexers which feed into one of four multimedia control centres. But,
although BVG has spent more than US$15 billion on new ATM devices,
Holger Seedorf, communications manager, says: "We are not replacing things
installed in the past, we are networking them with ATM technology. We are
integrating everything you find in a station- phones, destination indicators, PA
systems and emergency phones."
Siemens and BVG chose ATM because of its capability to transmit
high-quality video and audio: they wanted clear video pictures for real time
surveillance as well as retrospective analysis, and clear audio announcements
to passengers on the platform.
C O M M U N I C A T I O N Sc
I N T E R N A T I O N A L |
