It’s no secret that IP-based traffic has been moving onto service provider networks in droves in recent years. That has led these operators, many of whom have networks initially built for circuit-switched traffic, to seek a way to support both legacy TDM traffic as well as packets – and to do it in a way that involves provisioning and management that is somewhat familiar.
“What people were looking to do is take IP networks and simplify them so they could be managed more like traditional services,” notes Andrew Schmitt, directing analyst of optical at Infonetics Research
A few years back it was looking as if PBT (or, generically, PBB-TE) was the way the industry was going to go on this. BT (News - Alert), a pioneer in migrating to IP with its 21st Century Network, in the 2007-early 2008 time frame was very public in its endorsement of the PBT approach, which Nortel (News - Alert) was heavily promoting. Although BT’s excitement about PBT had dried up by mid-2008, when the operator said it wouldn’t be using the technology in the immediate future, PBT in the meantime built up a decent stable of supporters, at least on the optical supplier front.
But for all its early promise, PBT/PBB-TE has pretty much disappeared from the scene. Instead, a joint effort by the IETF and ITU-T called MPLS-TP is the packet optical transport standard around which the industry is coalescing.
“Many customers would like be able to set up Ethernet circuits in the same manner that they set up legacy SONET/SDH leased lines,” John Messenger: director of global standards activities at ADVA Optical Networking (News - Alert). “PBB-TE (802.1Qay) was developed to meet this need, but operator support for it has declined rapidly, and the majority of operators are looking toward MPLS-based solutions. To meet this need, MPLS must become more transport-oriented. In the same way that PBB-TE extended the existing Ethernet standards, we believe that a transport-friendly MPLS variant should be based on, and compatible with, existing MPLS standards, so defining it as a transport profile of MPLS makes sense.”
Indeed, that’s exactly what MPLS-TP, an MPLS extension or subset (as it is alternately described), aims to do.
“We think this is an incredibly important standard for helping merge together networks,” he adds. “This will be used on routers together with optical networks. It helps you provide a kind of end-to-end mechanism using MPLS, which is very well deployed in packet networks, in a connection-oriented manner. That means you can continue to provide the real high-quality services that operators still need to provision from their networks.”
Although the raison d’etre for MPLS-TP is to enable operators to manage packet-based connections as if they were nailed-up, circuit-based connections, just how exactly operations administration and maintenance will be addressed in the standard has yet to be settled.
Once that is agreed upon, says Smith, suppliers like Ericsson will begin rolling out MPLS-TP-compliant gear. He expects the first commercial implementations by service providers of MPLS-TP to go live this year, and to be turned up in quantity starting in 2011.
According to Smith, carriers are likely to use MPLS-TP first to support the bulk of their broadband traffic, including mobile broadband and DSLAM uplinks. “They may over time start to migrate the business services over to this part of the network, but another standard – the OTN standard – is helpful for that,” he says.
“What we don’t necessarily agree with is that using OTN to build packet networks is the right thing to do because ultimately it’s not going to scale,” says Luc Ceuppens, vice president of product marketing for Juniper's Infrastructure Products Group division.
According to Ceuppens, optical vendors are promoting the use of OTN, which he describes as a circuit-based technology that’s expensive, as part of their drive to MPLS-TP.
“For us, the whole thing around MPLS-TP is not necessarily around adding the extensions to MPLS as it is the drive of optical vendors to position OTN as the best technology to build packet-switched networks and also the best technology to take the cost out of the packet-switched network. We believe that is fundamentally untrue,” he says. “Every network we analyze, the moment you bring OTN into that network, the cost goes up and your complexity goes up.”
Nonetheless, Ceuppens says Juniper is actively involved with the MPLS-TP standard and intends to support MPLS-TP in its products. But, he adds: “We will continue to use and promote MPLS as the best infrastructure to deliver IP-based services because it is ultimately more adaptive to packet flow.”
Mike Capuano, director of marketing for service provider routing and switching at Cisco Systems Inc., a key player in the MPLS-TP standards effort, says “the important thing to note about MPLS-TP is it is effectively an extension of MPLS and therefore provides a seamless interface between the part of the network where you deploy MPLS-TP and the part of the network where you have what I would call dynamic MPLS deployed.
“That, we think, is important because over time it will provide the flexibility, if a service provider so desires, to [push] more dynamic MPLS out farther, closer to the end user,” says Capuano.
He adds that while delivering scalable MPLS is something that’s fairly easy to put on a data sheet, it’s fairly difficult to actually develop.
“A lot of folks talk about it; I think there’s really a very limited set of companies that can actually deliver on it,” he says. “Cisco’s obviously been doing MPLS for a long, long time, well over a decade, so we’re well positioned to deliver on that.”
However, if a service provider has a large team of technicians who are used to provisioning things manually, that operator may be more inclined to start with MPLS-TP. If a service provider already has a converged infrastructure for which it’s got a team of people that knows how to manage dynamic MPLS, that operator is more likely to go with a straight MPLS approach, which Capuano describes as dynamic MPLS (because it doesn’t rely on nailed-up connections as MPLS-TP would).
“The advantage to dynamic MPLS is it’s a self-managed network,” he says. “The network itself is intelligent so there’s less manual provisioning and calculating back-up paths and things like that.”
Whatever service providers decide is the best path to packet optical transport, however, Ericsson’s Smith says the IETF and ITU-T are jointly working to make sure MPLS-TP aligns as closely as possible with MPLS “so you can have some kind of seamless connection between MPLS and MPLS-TP actually in the network.”
Edited by Stefania Viscusi