MPLS MOVES DOWNSTREAM

It was once called a solution in search of a problem. Multiprotocol label switching, originally introduced as a way to speed up routers by giving them less to read (sort of the CliffsNotes of packet-based communications), became more popular as a means of traffic engineering that would ease the transition from highly precise ATM networks to highly permissive IP. IP/MPLS has become widely embraced in carrier backbone networks, but lately some have advocated pushing the technology further downstream, into access networks and even to the customer premises, sparking debate over where MPLS does and does not belong.

“We firmly believe that MPLS is moving closer and closer toward the customer, but it's a trend that's going to take some time,” said Lindsay Newell, vice president of marketing for Alcatel's IP routing group.

The rising popularity of IP video has buoyed the notion of using MPLS-based virtual private LAN services in metro networks for residential triple-play services. In addition, some carriers, such as SBC Communications, are using MPLS-based virtual private networks to replace legacy frame relay networks with IP networks that can carry voice and administer quality of service and traffic prioritization. Carriers are therefore using voice over IP to avoid cannibalizing their frame relay businesses. And although some large national corporations already use MPLS, it is trickling down-market.

“Smaller companies are saying, ‘We want to invest in MPLS technology,’” said Michael Kennedy, co-founder of consultancy Network Strategy Partners. “In that context, MPLS has become like a brand name, like Tide or something.”

In January, metro Ethernet equipment vendor Metrobility will begin to offer a new product that extends MPLS to the customer premises: a demarcation box with four access ports carrying tri-speed Ethernet (10 and 100 Mb/s and 1 Gb/s) and dual GigE uplinks that takes traffic from customer-facing interfaces and puts it into as many as 16 static label-switched paths. Because of its functions, it could be called a multi-tenant unit switch, said Manu Kaycee, Metrobility's director of system architecture, using a term given to an Internet Engineering Task Force specification.

Big carriers have begun to request such equipment to resolve issues related to cross-country metro Ethernet offerings, Kaycee said. For inter-city data services, carriers typically assign an Ethernet-based virtual LAN (VLAN) to each link in the access network, then supplant the VLAN tag with an MPLS tag in the core, reverting back to a VLAN at the other end. Using MPLS, carriers could simplify that process, using just one ID tag (or “name space”) on the traffic. With one consistent ID and no need for translations, there would be fewer errors, Kaycee said.

Carriers also could use MPLS's traffic management talents to perform end-to-end fault isolation. Before, carriers intending to offer residential triple-play services in the future had to be concerned because standards-based VLAN technology only supported less than 4100 tags. MPLS could give them millions, Kaycee said. However, an Institute of Electrical and Electronics Engineers draft standard for stacked VLANs would allow for millions of VLAN tags as well.

Without using MPLS end-to-end, vendors must find a way to better synchronize MPLS and Ethernet features. And standards-based solutions to that problem may be slow in coming, industry sources say, because of the divided jurisdictions of the various standards bodies. Ethernet is governed by the IEEE, MPLS is governed by the IETF, and neither wants to step on the other's turf. However, one IETF standard, HVPLS, seeks to create two different MPLS architectures — one in aggregation points and one at the edge of access networks. The Metro Ethernet Forum, meanwhile, is operating outside the two groups' territories, conducting an interoperability demonstration this month of various metro Ethernet access equipment with MPLS gear from the likes of Alcatel and Riverstone Networks. And the MFA Forum — the amalgam of the ATM Forum and the MPLS and Frame Relay Alliance — is focusing on the issue as well.

To some, however, pushing MPLS far outside core networks is hard to justify.

“It's pretty clear there's no significantly broad opportunity for MPLS, even within a metro area, certainly not all the way to the customer prem,” said Tom Nolle, president of consulting firm CIMI. Beyond 2010, MPLS may become useful in metro networks, he said, but not until then. “Vendors who are relying on that expectation are probably taking a pretty substantial, if not fatal, risk.”

As Nolle explains it, access networks are unlike the core networks that MPLS typically calls home in at least one important aspect: The vast majority of traffic in access networks travels predictably between just two endpoints — the customer premises and the exit ramp. Therefore, there's little need for a technology that routes multiprotocol traffic streams to their various destinations.

“There's not much point in putting IP infrastructure and MPLS features in to support what are going to be pretty static, consistent traffic flows that always go between the same places,” Nolle said.

Kaycee's response is that Metrobility isn't advocating MPLS in access networks as a means of creating efficient mesh networks; instead, MPLS is being tasked in this case with simplifying service provisioning and enabling end-to-end traffic engineering.

“The traffic pattern may be from the customer directly to the on-ramp, not varying, but what's the end-to-end service, and do we want some way of verifying the service end-to-end?” asked James Domine, vice president of product management for Riverstone Networks. “How does the traffic get from the customer to that on-ramp? Do we want redundant on-ramps, so customers have backup paths, and if so, how do we do that? MPLS gives us a reliability mechanism.”

Perhaps the two most prominent barriers to deploying MPLS anywhere it isn't already deployed are cost and complexity, though the fear of complexity has diminished somewhat as carriers have grown accustomed to MPLS. Kennedy argues that MPLS can be less complex than some existing solutions, as an MPLS-based broadband switch router can do the work of a digital cross-connect, an ATM switch and a router.

Kaycee claims his company is solving the price problem by designing its new equipment to be purpose-built for MPLS.

“Many of the existing devices have MPLS after the fact,” he said. “[Vendors] have taken ATM switches, Layer 2 switches or routers and added MPLS. None of them have been optimized. Our box is optimized to do really fast data-plane forwarding. As a result, the customer doesn't have to spend a lot of money on [dedicated light paths] or [border gateway protocol]. We don't run those protocols. [Metrobility's new equipment] is based on the VPLS drafts.”

Whereas carriers until now might have expected to pay around $25,000 initially to deploy MPLS gear, Kaycee said, Metrobility's gear will be less than $5000. Others say only the development of standards will truly lower the cost of MPLS access technology.

“We're definitely concerned about the cost of MPLS,” Domine said. Over time, he said, standards groups could alleviate the problem, but eventually, MPLS will be just one approach used by carriers.

“Every service provider has their own way of offering services,” Domine said. “Where I expect MPLS to be used, where we'd like it to be used, is where the carrier takes control of the service.”