Byline: JIM DUFFY
Here are just a few of the drivers behind Coca-Cola Enterprises' decision to roll out what it says - at 600 sites - will be the largest domestic network based on AT&T's Multi-protocol Label Switching technology:
* Lower operational expense.
* Improved quality of service.
* Support for IP telephony.
"MPLS is the reason we did it," says John Ridley, senior enterprise network architect at the $17 billion bottler, which migrated from a 6-year-old AT&T hub-and-spoke frame relay network.
But the Atlanta company is the exception instead of the rule when it comes to enterprise network adoption of MPLS-based services.
Carriers have plowed tens and even hundreds of millions of dollars into MPLS-related network upgrades to help consolidate facilities, slash operational costs and boost service reliability - all of which they say are panning out. But the service providers have been reluctant to aggressively market MPLS-enabled services to existing frame relay customers. Carriers say that service levels, security, reliability and pricing of MPLS-based IP VPNs have turned out to be comparable to that of frame relay - but not better.
Some carriers, such as AT&T, have a delicate balancing act between growing new MPLS-based services without cannibalizing revenue-rich frame relay services.
"It's just a question of where customers want to be," said Rose Klimovich, general manager of Managed Internet Access Services at AT&T, at the recent MPLScon conference in New York. "We're not going to force them to move from frame."
The next generation
MPLS is considered the next-generation carrier infrastructure and service enabler. It is designed to converge multiple networks into one that can accommodate any-to-any meshed connectivity between different access technologies, be they frame relay, IP, ATM, private line, DSL or Ethernet. It also is intended to enhance the performance, reliability and scalability of IP routing, making IP as dependable as ATM for data, voice and video traffic.
Virtually all facilities-based service providers are implementing MPLS in their core networks to improve efficiency and reduce costs. Many also plan to provision services, such as IP VPNs, using MPLS's ability to label packets and flows, and send them across explicit routes segregated from other traffic and networks.
AT&T will complete its next-generation converged IP network in 2005 with MPLS at its core. BellSouth recently announced its MPLS-enabled regional IP backbone, called BRIB. Verizon, SBC and Qwest also have MPLS buildouts in place or under way, as do Cable & Wireless and MCI.
Coca-Cola Enterprises is the beneficiary of these buildouts. The bottler is connecting its sites in an IETF RFC 2547bis-compliant VPN mesh that's based on AT&T's IP Enabled Frame Relay/ATM service.
RFC 2547bis specifies Layer 3 Border Gateway Protocol-based MPLS VPNs in which routing is usually outsourced to a service provider. And even though IP-enabled frame and ATM services utilize MPLS, observers say they are the same frame that ATM VPN customers have been using, only optimized for IP.
"We are consolidating all of our data centers into one location in Atlanta," Ridley says. "We will now have to support application users all the way from Brussels to Hawaii out of the same data center. To do that, we need to be able to have a consistent end-user experience for those applications, and they're traversing far longer distance than they used to. So we needed a way to control congestion inside the carrier cloud."
The most practical way to do that is with Differentiated Services quality of service (QoS) over an MPLS network, he says.
Coca-Cola Enterprises' network in the U.S. will support ATM and frame relay local access, and regional data centers interconnected via OC-3 ATM virtual circuits. Remote circuits are fractional T-1 and fractional DS-3.
The company's European network was similar to the frame relay network in the U.S., with offices hubbed into the U.K. In Europe, the bottler is migrating to a British Telecom MPLS network with Layer 2 services provided by frame relay instead of ATM, Ridley says.
"There were three or four good reasons for doing this," Ridley says. "We wanted the any-to-any connectivity [MPLS provides]. As we lifted and shifted out of our regional data centers and put things into our central data center, we didn't have to worry about shifting traffic patterns."
The bottler also wanted "end-to-end" QoS to ensure that applications such as Citrix would "survive" as data was transmitted, for example, between Atlanta and Lyon, France, he says. Coca-Cola Enterprises also wanted a network that could evolve to support packet telephony.
"And as time goes on there will be other latency-sensitive applications that come along, like streaming video and audio," Ridley says.
Another company, software maker MoldFlow in Wayland, Mass., switched to an MPLS-based IP VPN after its frame contract was up, though it was the price of the new service rather than the fact it was based on MPLS that grabbed IT Manager Rick Thimble's attention.
MoldFlow isn't in a position to exploit the traffic engineering and QoS capabilities of MPLS, Thimble says. "It's basically e-mail that we're putting over the lines," he says.
MoldFlow switched to a fully managed Cable&Wireless IP VPN (www.nwfusion.com, DocFinder: 6135) when it was time to renew its multinational frame relay contract with AT&T, which had just raised its prices on low-speed circuits (DocFinder: 6136). MoldFlow explored different service options with AT&T, including its IP Enabled Frame Relay service, but that would have cost more than a new frame contract, Thimble says. The C&W IP VPN service is half the cost of the AT&T frame service MoldFlow was using, he adds.
A tough sell?
Carriers acknowledge that most enterprise network customers don't care what enables a service as long as they receive their contracted service-level guarantees.
And if frame fits that bill, carriers say there are no compelling reasons to encourage a migration to MPLS-based IP VPNs.
"They need an easy way to come in and move to MPLS VPNs over time," AT&T's Klimovich says. "We expect some to stay on frame."
"Frame and ATM are still alive and kicking," said Mark Logan, head of VPN products for British Telecom Global Services, who also spoke at MPLScon. "But not like it once was."
Carriers say new data network customers will be steered toward MPLS-enabled IP VPNs because that's where carrier networks are headed. Customers with many sites to connect, applications to converge, extranets to construct, class of service to enable, operational costs to reduce, and current frame contracts that expire will be encouraged to evaluate the new services and technology.
British Telecom will market its MPLS IP VPNs as a replacement for competitors' frame relay services. But Logan says pricing for the service will be comparable to frame relay.
"The carrot's been the money saver, but the pricing differential is not that significant," says Rosemary Cochran, principal at research firm Vertical Systems Group. "So there's got to be something else."
There is, Klimovich said. There's a potential cost reduction in site connectivity for businesses with a lot of locations. Users can establish extranets for customer and supplier interaction. They also can network all their applications over one infrastructure.
MPLS can be a platform or foundation for new technologies, she said, and reliability has been improved over the last two years.
"MPLS IP VPNs are ready for prime time," Klimovich said.
At MPLScon, AT&T would not disclose how it is pricing MPLS IP VPN services compared with frame, but in the past the carrier indicated the pricing structures would be similar (DocFinder: 6137). The company also could not predict when its MPLS IP VPN revenue would overtake its frame revenue. Klimovich said there is "pretty significant growth" in AT&T's MPLS IP VPN services, but also growth in its frame business.
For British Telecom, Layer 2 VPNs based on frame and ATM still account for 70% of its VPN revenue, but it could see MPLS IP VPN revenue overtake frame in two to three years, Logan says. That's consistent with IDC's 2006 forecast for Western Europe.
Other analysts say they think it will take longer.
"That's not a two- to three-year time frame, that's a five- to seven-year time frame," says Mark Bieberich of The Yankee Group.
"I hope to be retired" by the time VPN service revenue overtakes frame, Vertical's Cochran says. Worldwide frame relay service revenue in 2002 was $14.5 billion, while revenue for dedicated IP VPNs was a fraction of that - $791 million, she says.
A recent Vertical survey of 563 enterprise network managers showed that 16% of respondents planned to convert to a dedicated access VPN. Moreover, customers planning to switch represent only 3% of the 42,000-plus total frame relay locations reported for all respondents. That means the average size of these networks is relatively small.
But, like Coca-Cola Enterprises, a few large companies are blazing the MPLS trail. Some are implementing MPLS in their private networks for tactical reasons, such as isolating business units with RFC 2547; consolidating addresses of merged or acquired companies; and to divide traffic loads among links, says Bruce Davie, a Cisco fellow at Cisco.
Other customers inquire about MPLS because it's trendy, carriers say.
"It's kind of a hot thing; they are aware of it," says Stuart Elby, vice president of network architecture for Verizon enterprise technologies. "It is somewhat of a selling point."
"I think there's a lot of that fashion stuff around MPLS VPNs," British Telecom's Logan says. "There's really no logic to it. Yet 20,000 lemmings can't be wrong."
Unless an existing frame contract is in place, everything's working and nothing needs to change. In that case, "it's not an easy sell," Cochran says.

Комментариев нет:
Отправить комментарий