ATM - life in the old dog yet

June/July 2002

With the communication industries continuing to embrace IP as the lowest common denominator for multiservice network transmission, some observers believe that the days of ATM - the traditional mulitservice networking technology - may be numbered. While ultimately, this may be true, news of its imminent demise has been greatly exaggerated.

In the broadband world, it is not just data that moves fast. As recently as 1999, ATM (asynchronous transfer mode) was being hailed as the ‘next big thing’. Just three years on, ATM is derided as an expensive dinosaur in some circles, soon to be replaced by cheaper and simpler IP-based networks.

ATM has significant limitations in terms of cost, complexity and scalability, claims Alan Taylor, European technical director of Juniper Networks. Juniper, the second largest vendor of Internet Protocol (IP) based routers after Cisco, argues that carriers and service providers are in the process of ripping out ATM core networks and replacing them with cheaper, service-based IP networks. ‘IP offers better performance at a fraction of the cost,’ says Taylor.

However, according to Andrew Rufener, Marconi’s director of technology and an ambassador for the ATM Forum, ATM is going to be around for a long, long time. Rufener believes that companies are actually expanding ATM networks because it is the only technology that will help them offer potentially lucrative 3G and DSL services.

But, as with most technology arguments, the truth actually lies somewhere in the middle. ATM will eventually be replaced by IP-based networks, which offer broadband providers a fast and cost-efficient network that it is flexible enough to support a range of value-added services such as managed VPNs, networked applications and rented applications. ‘MPLS (Multiprotocol Label Switching) is clearly the wave of the future for carrier core networks, representing a good compromise between IP’s flexibility and the desire for more performance to support applications like voice and video,’ says Jim Slaby, a research director with Giga Information Group.

Yet while sales of IP infrastructure are growing rapidly among next-generation companies, Giga recommends companies wait for at least 18 months before making the leap. Many incumbents say they are unlikely to switch to IP until their current architecture is fully outdated, which could take ten years or longer. In the meantime, it is likely that ATM and IP will become complementary - rather than competing - technologies.

Pioneering

When it was first introduced in the early 1990s, ATM was the first networking technology to combine the efficiency of traditional packet-based networking with the speed and reliability of connection-based networking. ATM works by combining voice and data into single packets called ‘cells’, which are then sent along a pre-defined network route at speeds of up to 622Mbps. ATM broke new ground because it allowed video and audio to be sent across busy wide area networks and arrive at the other end in intelligible form.

In the last ten years, companies have invested billions in ATM networking and it is now a de facto standard for most carriers, service providers and enterprises, says Lisa Pierce, a research fellow with Giga Information Group. Although ATM is increasingly being pressured by newer networking technologies such as IP and Gigabit Ethernet, few companies are likely to ditch their existing ATM infrastructure, Pierce says.

‘Banks shouldn’t be using leased lines because there are cheaper alternatives, but they still are - and the same will happen with ATM,’ she says. ‘Main carriers and ILECs (incumbent local exchange carriers) have made a massive investment and they are not going to want to go to the expense of ripping out systems and starting again in the current climate.’

Much of the 70% growth in ATM spending has come from service providers that are dependent on the guaranteed QoS that ATM provides, says Tim Forshaw, marketing director with Alcatel Broadband. ATM can guarantee that a cell stream will arrive on time and making sense. As a connectionless technology, IP cannot match this - data packets are split up and sent separately across the Internet using whatever route is available.

In terms of technical capability, there is little to choose between ATM and IP, particularly since the launch of 10Gbit ATM routers and switches. However, the industry is slowly moving towards IP because it offers a cheaper and less complex platform for deploying services, believes Peter Aknai, a senior consultant with Analysys.
‘Because of the heavy discounting of IP products, it’s around twice as expensive to send data across an ATM network compared to an IP network,’ Aknai says. ‘ATM is also becoming more expensive to maintain as vendors drop ATM product lines and support packages.’

Complex…

ATM’s connection-based networking could also be a headache to maintain as networks become more complex, explains Forshaw. In an ATM network, each connection has to be configured and managed with a series of hubs, routers and switches. In an IP network, there is no need for this complexity, since the network is connectionless.
‘If you have to manage a point-to-point connection, it becomes very difficult to roll out multipoint systems like webcasting,’ says Forshaw. ‘These kinds of new services are key ways to generate revenues, but will require an alternative transport mechanism.’

IP infrastructure products are also making important steps towards matching the reliability of ATM. The introduction of MPLS functionality allows service providers to prioritise data packets and implement queuing systems. MPLS, based on Cisco’s proprietary tag switching protocol, is a way of provisioning QoS and works by labelling each packet as it passes through the edge router. The labels can be assigned different priorities so that video data is prioritised during working hours, for example, or particular customers always receive fast service.

However, MPLS is a relatively immature technology and until standards are more clearly defined IP services will not be an option for mainstream companies, believes Giga’s Slaby. ‘Until MPLS is an industry standard rather than a Cisco standard, any investment in MPLS and IP puts the company at risk of being stuck with obsolete products or paying for costly upgrades two years down the line.’

Making matters even more complicated, Slaby points out that no fewer than four different bodies are arguing over standards for IP services. ‘Until the Internet Engineering Task Force (IETF) makes a decision, the issue won’t be settled. This isn’t something that will be sorted out in the next month or two, put it that way.’

…though dependable

With IP still not ready for primetime, most enterprises will turn to ATM as a support for the newer technology. An emerging category of ATM switches known as multiservice WAN switches deliver both Frame Relay and IP, as well as ATM, although ATM is the underlying transport mechanism for all those services. Companies spent $4.3 billion on these switches in 2001, according to Yankee Group, and spending is predicted to reach $10.6 billion by 2005.

Alcatel has already sold 450 of these platforms to customers, says Forshaw. ‘You can have your ATM infrastructure, with the reliability that involves, and we can migrate other services like IP onto that infrastructure,’ he says. ‘So the customer buys Frame Relay or Cell Relay from their provider, but they don’t care where it comes from.’

Most companies will rely on these products to support broadband IP applications on an ATM infrastructure for at least five years, Pierce adds. ‘New players - and there aren’t many of those - will roll out IP in place of ATM, but most existing companies will use ATM as the chassis on which everything else - voice, video and data - will run,’ says Pierce.

That is the model at BT Ignite, which takes the multiple services that a business or home might subscribe to and integrates them on a single network connection delivered through the BT network, cutting down substantially on cost because it uses a single network fabric. As a result, the network can offer QoS on each application as well as dynamic bandwidth allocation.

‘We see ATM and IP absolutely as complementary technologies,’ says Helen Stiles, BT Ignite’s corporate VPN marketing manager. ‘You can run an IP application over an ATM infrastructure, so customers can migrate easily from Cell Stream services. ATM will support pretty much any application or technology a customer wants to run.’

The BT Ignite package of services includes ATM, IP, Leased Lines and Frame Relay, with a Gigabit Ethernet service launching later this summer. ‘There is certainly still a lot of demand for ATM, but we find that customers increasingly have a combination of technologies,’ says Stiles. ‘What we offer is a risk free way to have the best of all worlds - we will use whatever underlying technology is best suited to the customer’s needs.’

It is a business model that is likely to be found across the industry, says Aknai. ATM will remain the technology of choice for backbone and incumbents, with pure IP networks limited to enterprise Internet communications and start-up providers. However, ATM is likely to be exploited as a way of delivering IP services to customers in most cases. ‘History teaches us that we rarely abandon a technology but simply keep it going alongside the new ones,’ says Aknai. Either way, we certainly have not seen the end of ATM.