Hey, you, get through my cloud

Multiprotocol label switching, commonly known as MPLS, is going to make it easy, affordable and convenient to zip packets across data networks without caring what makes those networks tick or, better yet, whether those networks understand the kind of packets they're transporting.

MPLS is a way to speed IP-based data communications over ATM networks. An MPLS router at the edge of a network cloud reads the packet information, applies a tag and sends the packet to switches in the cloud's core, which then route the packet to the right destination on the other side.

Because the packet has been tagged via MPLS, there's no need for the core switches to open and examine the packet's contents; thus, it moves rapidly through the network, improving quality of service (QOS).

That's the theory behind this nonstandardized procedure that is gaining acceptance within the vendor community.

“The vendors who support MPLS include folks like Cisco and Juniper,” said Prabhu Kavi, product planning director for Tenor Networks. Cisco Systems, which developed a version of tag switching, is widely credited with starting the MPLS ball rolling.

MPLS is being adopted by some early users such as Global Crossing and Cable & Wireless and “a large number of other folks who use it internally but have not made an announcement about it,” Kavi said.

Part of the reason MPLS-enabled equipment is not being embraced like a cuddly newborn is because it has an ugly birthmark: MPLS is not standardized. Without standards development, costs could be wasted if things go another way. This is especially painful for vendors that build hardware.

“MPLS has gone through a long road map,” said Sam Halabi, vice president of IP carrier marketing and business development for Extreme Networks. Halabi sits on the board of the MPLS Forum and is confident that the winding road has been enough to at least encourage initial deployments. However, “that doesn't mean that MPLS itself is not a moving target,” he said.

Couple that with an unsteady economy, and it's understandable that there's more MPLS buzz among vendors than users.

“I see [MPLS] slowed down by the current economy because people who would rush out and say, ‘Yes, I want the latest and greatest’ are having to review and say, ‘What do I really need at this point to keep my current computing capabilities up with my competition?’” said Ron Mohr, senior marketing manager for Alcatel's PowerRail product line.

Vague customer interest, though, won't stifle development. “We still have to create this stuff. It hasn't slowed down for us,” Mohr said.

Vendors don't want to be left out when a new technology takes hold, and they definitely don't want their gear to be stranded if the network architecture shifts and they don't. On the other hand, there's that moving target issue.

In some instances, MPLS even conflicts with what's already out there — or at least what's proposed.

One MPLS selling point is its ability to deliver QOS across the network. This is an important part of cable's DOCSIS 1.1 packetized data standardization effort. Thus, a high-speed broadband data provider must be cautious about how MPLS fits with its next-generation cable modem products and determine whether incorporating MPLS capabilities produces overkill.

On the other hand, other benefits of MPLS don't relate to QOS and fill in holes left by DOCSIS 1.1, said Elisa Camahort, product marketing director of Terayon Communication Systems' cable data headend solutions group. Terayon is in the process of incorporating MPLS into its cable modem network products.

For example, MPLS tops DOCSIS by making packet processing more efficient and improving traffic engineering. MPLS, Camahort said, “can happen across ATM, across the hybrid fiber/coax network or eventually, they're working on having it work across dense wave division multiplexing. It streamlines the packet processing and removes some of the processing power required.”

For traffic engineering, MPLS looks for the most efficient path so all packets are routed not just on the shortest path, but also on the most efficient one.

Even so, MPLS doesn't do enough, said Opher Kahane, president and CEO of Kagoor.

“MPLS is a generic technology,” Kahane said. “It does not deal with application-specific needs — in this case, voice.”

Kahane admires MPLS and said that Kagoor's proprietary traffic management technology will interface with — or “complement” — MPLS when the time is right, but for now, MPLS is not the cureall its supporters claim.

“There's so much buzz going on out there in the market regarding MPLS,” Kahane said. “There are things it does well and things it doesn't do well, and effectively dealing with voice-over-IP is something it comes short of.”

In fairness, MPLS was not intended to deal specifically with voice over IP (VoIP). It's a data transport mechanism. “MPLS's killer application was found about two years ago. That was traffic engineering,” Kavi said.

‘As long as there are legacy protocols and legacy networking capabilities and there's over $1 trillion in investment out there, this isn't going to disappear overnight and get converted to MPLS.’
— Marco Pagani, Nortel

He believes MPLS can handle VoIP because it's engineered to bring QOS to the network. “MPLS can be pretty important for something like IP telephony,” Kavi said.

Nevertheless, VoIP is not the target application for MPLS, he said.

“Where this really makes sense is where bandwidth is expensive,” Kavi said. “When it comes to things like the long haul, that's where it makes a lot of sense. And [our product] actually acts as a high-speed packet switch that understands multiple classes of service.”

Alternatively, the killer app could be transporting packets across a metropolitan area network, said Alcatel's Mohr.

Users “can have local area networks on different sides of the city and utilizing the [network] cloud with MPLS, they can have control over more than one local area network, but the system sees it as one,” he said.

This, Mohr said, is especially important with virtual local area networks (VLANs). “We can keep that VLAN tag on the packet as it goes through the cloud so when it gets to the other side it acts as if it came from the same VLAN,” he said.

That's if people ever decide to use MPLS at all.

“I believe the demand will be there,” Mohr said. “Right now I think people see it as a neat toy. As speeds on the edge increase, and the need to utilize the cloud grows, nobody on the edge wants to worry about what's in the middle. If I send a packet out, it should get across, and I think MPLS is going to be a great resource for that.”

But that's only if the packet doesn't contain voice, Kahane said.

“It's clear that the underlying IP technology, which is so effective for data applications, is simply too generic, too agnostic to effectively deal with the real-time complexities of voice and fax traffic,” he said.

That could be why nearly everyone agrees with Marco Pagani, vice president and general manager of Nortel Networks' carrier networks division. Nortel is building products that will be the building blocks of the next-generation IP core, which will be MPLS-centric, he said.

The catch is when that network will arrive. “As long as there are legacy protocols and legacy networking capabilities and there's over $1 trillion in investment out there, this isn't going to disappear overnight and get converted to MPLS,” Pagani said.