Which protocol will dominate carrier data networks by 2010?
Several months ago, visitors to Lightwave Online were greeted with the following question: "Which protocol will dominate carrier data networks in 2010: IP, Ethernet, ATM, frame relay, or something else entirely?" Our online pundits clearly favored the first two alternatives. However, when we asked a few industry experts the same question, most found it too simple to take into account the key variables that will determine protocol choice among applications. Geographic location within the network (metro, core, or access), which layer of the protocol stack is under discussion, and the soapbox from which the opinion is being delivered all can color the answer to this question. The current economic malaise also makes it difficult to predict what will happen next year, let alone seven years from now.
That said, simplifying the network to the point where there is only one protocol seems to be the goal. The ideal is "IP over photons," asserts Marlis Humphrey, chairwoman of the board of directors of the ATM Forum and director of technology and standards planning for Harris (Melbourne, FL). "That's the Nirvana. That's where everyone is trying to go, and that's probably the essence of Lightwave's question about 2010: Can you really get to just IP over photons?"
The answer, say our sources, is that it's not likely—at least in the near-term. Humphrey knows of many enterprises and service providers that today run IP over ATM over SONET over photons, for example. "It sounds like a lot, but that is the only way they can achieve what they want to do, which is the efficient carriage of data services," she explains. "They would like to start pulling out those excess layers and getting to a more efficient network and a more collapsed protocol stack, but they just can't get there at this point in time. Each piece has its value in the network, and it's going to take some time before we can evolve to what might be an idealistic view of one protocol."
Even if we conclude that one protocol would be ideal, not everyone agrees with Humphrey that IP will fill that role. Given the lack of consensus on what the target will be, carriers that need to make their networks more data-friendly now can be forgiven if they are ruled more by expediency than an industry-wide vision. Which may mean that tomorrow's data network will look very much like today's.
Certainly, there is no shortage of people who believe that change is coming. "It's clear now that in the next 10 years, Ethernet will become the overwhelmingly dominant protocol in metro networks," asserts Michael Howard, principle analyst and co-founder of Infonetics Research (San Jose, CA). "And it's not too much of a leap of faith to think that in 2010, we're going to see at least some long-haul routes using Ethernet as the main protocol."
Ethernet is ubiquitous in the LAN, thanks to its simplicity and cost-effectiveness. "It is not necessarily the best protocol compared to many others in the market," admits Nan Chen, president of the Metro Ethernet Forum and director of product marketing at Atrica (Santa Clara, CA), "but Ethernet has been winning in the enterprise space for the past 20 years because it's the cheapest 'good enough' technology or protocol to win."
There are critical resiliency and quality of service (QoS) hurdles to be overcome before Ethernet becomes a mainstay in the metro, however, including how to guarantee delivery of real time delay-sensitive traffic over Ethernet. It's an efficient and effective carrier of data, but how can you optimize Ethernet for bursty voice and video services? "It's one thing to create a large Ethernet network," notes Rick Townsend, president of the ATM Forum. "It's quite another to say, 'I am part of the public network.'"
First, says Chen, a set of standardized Ethernet services must be defined, so customers who buy services will know what they will get. Such definitions will also serve as a common denominator for vendors that are building the equipment and carriers that would rather worry about the services than which or whose equipment to use. Second, Ethernet must be made carrier class, with guaranteed service delivery rather than just best effort. The Metro Ethernet Forum is currently working to satisfy these needs.
"From the services perspective, we define Ethernet line services and Ethernet LAN services," explains Chen. "Ethernet line service has been defined as a point-to-point service, and Ethernet LAN service is a multipoint. Those two services will provide the technical foundation to enable service providers to offer a myriad of services." Chan predicts that these definitions should be completed by August.
The forum has also defined what it calls committed information rate (CIR) and peak information rate (PIR), whereby CIR is guaranteed bandwidth and PIR is best effort bandwidth. "There could be a service definition where you had 50 Mbits/sec of committed information rate and 100 Mbits/sec of bursty information rate," explains Chen. "Regardless of network traffic—busy or not busy, congested or not congested—that CIR is always guaranteed at 50 Mbits/sec and the other is available as the network is more available." The forum believes that will make Ethernet more appealing to end users who need dedicated bandwidth for video type of applications, for example.
Assuming Ethernet overcomes QoS and reliability issues to dominate the metro, just how likely is it to become a major contender in the core? "We saw the LANs go to Ethernet, and I believe we can see enough to know the metro is going to go to Ethernet, so why leave out the long-haul?" asks Howard. "It's just interconnecting the metros, and it's the final piece to put in place."
There is no pressure as yet to migrate Ethernet into the long-haul, however. DWDM is relatively inexpensive, says Howard, and there's plenty of it installed. Given the amount of available fiber, it's easier to keep adding to today's DWDM networks. But 10 years from now, who knows?
"The service providers will be offering end-to-end services," reiterates Howard, "and since there is probably nothing that is going to displace Ethernet at all these business locations and even homes, I think Ethernet is going to be the main [protocol]. But MPLS may be an important protocol to make it carrier class all the way into homes and businesses even."
MPLS can be regarded as the enhancement and evolution of IP routing, explains Gary Leonard, marketing awareness and education committee chair of the MPLS Forum and director of field marketing at Riverstone Networks (Santa Clara, CA). IP is connectionless and does not provide the QoS features of a circuit-based service like ATM or FR, and that's where MPLS comes into play. It separates the control function from the forwarding function of an IP packet by appending a 32-bit label on the outside of the IP packet, which helps MPLS-enabled routers to determine the most efficient route through the network.
A key advantage of MPLS is that it can complement Ethernet, ATM, FR, or packet over SONET/SDH. "MPLS is not just the evolution of IP—although it is," explains Leonard. "And it's not just traffic engineering in the core—although it is. It's now an access technology and a legacy migration technology. It's being used for creating the carrier networks' value-added services."
Roger Ruby, president of the Frame Relay Forum, also tosses his hat into the MPLS ring. "MPLS has a lot of potential to take the legacy traffic today and be able to encapsulate it and put it over a common core," he asserts. "There is a lot of work being done to allow for differentiated services and quality of service, which points us toward that 'Holy Grail' of a converged network that can carry voice, video, and data. I think MPLS has the most promise in that area."
Infonetics's Howard contends that ATM and frame relay (FR) networks could disappear entirely by 2010. After all, the Metro Ethernet Forum is working to define Ethernet to have the same resiliency and guaranteed bit rate as these circuit-based technologies. However, he adds, "ATM and frame relay as services are going to last longer than the ATM and frame relay networks will, because the actual network will become an MPLS network."
But lately, ATM has seen its popularity increase as a result of current economic conditions. "From our experience, ATM seems to be selling extremely well into networks," reports Humphrey of the ATM Forum. "There was a point at which we said, 'Okay, it's going to start eroding from the core. We'll see it only at the edge, and that is going to happen really fast.' This was about two years ago. The economic downturn has caused people to be more concerned with business imperatives, and ATM has enjoyed a resurgence in popularity to some extent because it still makes money."
According to a recent Yankee Group (Boston) report, ATM business services will experience a 9.4% compound annual growth rate over the next five years. However, the research firm also notes that while ATM continues to enjoy a large presence in carrier networks, the carriers are slowly migrating away from ATM toward IP.
In the near-term, though, carriers will continue to rely on legacy services. Sprint revealed during its last quarterly earnings announcement that its long-distance services have stabilized. The carrier saw 20% revenue gains for FR and ATM. "Moreover," says Humphrey, "the margins on those services are much higher. It makes business sense still today. Even by 2010, ATM won't be entirely ripped out."
The ATM Forum's Townsend agrees. "The RBOCs and service providers are only buying equipment to meet demand, and in that kind of environment, they're going to buy something they already have," he says. "They can put it in, and their people know how to work with it; they know how to manage it. To start with a new technology today is an uphill battle."
The Frame Relay Forum's Ruby believes that the technologies of today, such as FR, for example, will be reutilized as access technologies. The core will be something other than FR, but the service will remain. "We will still see the need for some kind of private network, a closed secure type of service that can't be hacked—that's one of the main benefits that frame relay can provide," he says.
According to market researcher and strategic consultant Vertical Systems Group (Westwood, MA), FR services are also in demand; the worldwide market should reach $21 billion by 2004, up from $11 billion in 2000. In the United States alone, FR revenue will reach nearly $12 billion by 2004, driven by the expansion of existing customer networks, the demand for higher-speed ports, new customer installations, and global connectivity. Like ATM, however, FR will be increasingly challenged by emerging technologies and access options, including dedicated virtual private networks, Gigabit Ethernet, and DSL.
Of course, vendors and analysts can speculate about emerging protocols and the future dominance of Ethernet, MPLS, and IP all they want; it's the people with the money who count, and they tend to be a pretty conservative bunch. As the market data indicates, technology that brings in revenue will be very difficult to displace from the network. Which means you might be able to see tomorrow's network today.
"I don't think you'll see a lot of change between what's out there now and what will be out there in 2010," observes Steven Wright, principal member of the technical staff in the Advanced Network Architecture Concepts Group of BellSouth (Atlanta). "It's only seven years away, and that's not enough time for a major build-out of anything new in terms of wide-scale deployment. There will be continued progress in IP/MPLS in the cores and maybe a bit more Ethernet around the edge, but there's still going to be a lot of SONET transport out there."
As for the demise of ATM and FR, Wright says that he "struggles to recall the last time [BellSouth] removed a tariff. In that sense, the services will be around as long as people are interested in them. As always, what's in the core or behind those services depends on what is the most economical and efficient way to carry them."
Perhaps, as Wright asserts, "It's just a matter of evolving slowly from where we are. Given the current economic situation, I don't think there will be radical changes in the near-term."