Word of ATM's pending death was greatly exaggerated. Any number of analysts predicted that the use of Asynchronous Transfer Mode would die out somewhere within the last two years, but for several reasons, this legacy still has a stronghold in the network edge, and may be gaining in public networks. Why is this, and where will things stand in the future? Linda Dailey Paulson has the story.
ATM is a networking technology with a strong installed base. People know what it is and how it works. It's a known quantity. There's reluctance to discard it, and with good reason. New services are being offered via ATM networks, including voice, DSL, Internet access, video; the advent of Multi-Protocol Label Switching (MLPS) is also providing more opportunities and challenges.
Word of ATM's pending death was greatly exaggerated.
"Eighteen months to two years ago we thought ATM would die out or down," says Christin Flynn, analyst for The Yankee Group. "It really hasn't." She says there has been a surge of growth in ATM as what she calls a "back haul technology," precipitated by the growth of frame relay. "There's a lot of it and we're still seeing healthy growth."
According to Flynn, ATM switching had a 50 percent growth between 1999 and 2000 with the market for multiservice WAN switches at $4.3 billion in 2000. Traditionally, the firm has examined ATM only, but frame relay through multiservice WAN was added to the survey with the growth in these services. Multiswitch vendors include Alcatel, Cisco, Ericsson, Lucent, Marconi, Nortel and WaveSmith Networks.
She expects that market to have a compound annual growth rate of more than 20 percent during the next five years, reaching more than $10.6 billion in 2005.
"ATM is still today the only way to handle mixed services traffic cost-effectively where quality of service is paramount," says Marlis Humphrey, chairman of the Board of Directors of The ATM Forum.
Although sales were sluggish in 2001, there will be growth in the market attributable to the backlog of carriers that had been waiting for either MLPS or a higher-capacity switch fabric and density, and to these carriers resuming their purchase of high-density core devices, Flynn says. Growth can also be attributed to increased frame relay services and traffic.
"There's a big push for ATM in the public network," says Rick Townsend, president of The ATM Forum and a distinguished member of the technical staff at Lucent Technologies. "The exciting stuff is happening in the public network with carriers and service providers. It begins to feed back into the metro area networks."
Where ATM has a stronghold is at the network edge. "A year ago what people thought would completely go away was pushed to the edge of the network," says Agilent's Trevor Dyck, a product manager for RouterTester. "We've seen a rebirth, a refocus on ATM at the core." With about 17 billion USD invested in ATM and frame relay, he says, there is not going to be a "switch to IP or MLPS overnight."
What is happening is a move to hardware that mixes these transport technologies; typically ATM with pure IP or MLPS. Much of the reason ATM remains a preferred technology is the quality of service it offers when compared with technologies such as Ethernet. There's a lot of talk about how big MLPS will indeed be in the future, but its advent does not mean the end of ATM.
"The question is when will there be a substitute technology either in the enterprise or public network,"' posits Humphrey. Although MLPS is a "step toward that," ATM remains a critical layer within networks -- public and private -- and will remain so for maybe even 10 years. This in great part due to the added control characteristics it offers in enterprise settings and its quality of service, particularly for mission critical uses.
Flynn says the current "ideal is to take ATM and put it over existing MLPS from the core to the network." Although MLPS is still a young technology, "I think will happen in the core of the network by 2003,' she says. "That's when you'll start to see more ATM over MLPS in the core."
"I think there is a perception MLPS will replace ATM," says Dyck. In fact, there is interest in "MLPS technology that allows you to run both ATM and native IP in the network. MLPS own its own does not solve quality of service issues. It still has to be used with QoS mechanisms."
Despite the additional perception that there is a push in the enterprise to replace ATM with IP, network administrators are expected to retain their legacy investments in ATM and will transition over time to IP. The big interest now seems to be in hardware that retains the existing ATM infrastructure while allowing IP to be layered on top of it.
"Even once a substitute technology is developed ATM will continue to evolve," says Humphrey. This has been the case throughout the life-span of this technology. "When everyone thinks ATM everyone thinks cells ... but ATM has changed. There are different versions of ATM ... and it is changing over time as IP is. IP looks nothing like it did 10 to 15 years ago."
Humphrey says ATM is now able to transmit either fixed length or variable packets.
Most advantageous to ATM's future is its strong base. "Even when there is an alternative technology," says Humphrey, "the transition costs are extensive. It's not just new equipment. It's an issue of training people and of mitigating the risks of moving to a new technology. The longer it takes MLPS or IP over lightwaves to be evolved and deployed, the more entrenched ATM networks become."
Another reason enterprises are hanging on to ATM is their legacy applications, particularly for secure transactions. Although new transport technologies might be deployed within the network, administrators are loathe to give it up and risk security breaches that might be possible with, say, IP over VPN.
Flynn says different enterprises have differing needs and ATM is important in many different vertical markets such as schools, hospitals and government. "They use ATM more and are continuing to deploy ATM. They have traffic that is mission-critical," she says. These enterprises need both speed and security. "Having that reliability, that quality of service is very critical to some of these types of enterprises." One such example being telemedicine.
Townsend says this is true as well in large corporate networks where there is a large ATM investment. "There's no reason to change," he says. "Once they've got it they're not going to give it up. It doesn't offer that much in savings to take it out and put something new in."
One factor folks had not counted on as regards ATM's continued stronghold was the impact the national economy compounded by the events of September 11 had on network deployment and investment. "It slowed innovation worldwide," says Humphrey. "This is something we could not have predicted."
Linda Dailey Paulson