Some believe MPLS, a core networking technology today, is ready to go and ideally suited for the job. Others suggest a stripped-down version of MPLS, called T-MPLS, is the best way to bring MPLS into metro/aggregation networks in an affordable and manageable way. Meanwhile, BT and Nortel Networks Ltd. want to add PBT, or provider backbone transport, to carrier Ethernet switching products in an effort to allow for better traffic engineering on IP transport networks.
T-MPLS
The T in T-MPLS stands for transport, as T-MPLS is a subset of MPLS specifically designed for transport applications. The idea here is to simplify MPLS from an OAM&P standpoint, explains Mike O’Malley, group manager of portfolio marketing at Tellabs Inc.
 Tellabs’ Mike O’Malley |
Tellabs was among the companies that initiated work on T-MPLS within the ITU back in May 2005. However, it is Alcatel that is generally considered the key proponent of T-MPLS. Alcatel declined to be interviewed for this article. Alcatel spokeswoman Tracy Dupree indicated to xchange in December that the company was reviewing its strategy in light of the merger with Lucent Technologies Inc.
Although T-MPLS has made strides on the standards front at the behest of several important vendors, no service providers publicly have endorsed T-MPLS. However, O’Malley says there’s a lot of interest in T-MPLS from services providers, but he adds that Tellabs also is looking into PBT and has no preference for how the market goes on this one.
Architecture approval for T-MPLS — which is further along than PBT on the standards track — was granted within the ITU in November, so the first set of recommendations for this technology is complete. Additional work on OAM extensions for T-MPLS is ongoing and is targeted for consent in June, according to Tellabs’ spokeswoman Ariana Nikitas.
“If you look at the full list of MPLS commands — label swapping, etc. — there are a lot of things that are in MPLS that are meant for router updates and things like that that don’t pertain to transport equipment,” says O’Malley. “The idea behind T-MPLS is to use that goodness of MPLS from an end-to-end management and traffic-engineering standpoint to manage multiple pipes of transport equipment similarly, but to do it with a reduced command set.”
PBT
By most accounts, PBT is basically just another way to accomplish what T-MPLS is trying to do.
 Nortel’s John Hawkins |
 Juniper’s Kireeti Kompella |
PBT “provides a means of traffic engineering the network so paths through the Ethernet cloud can be pre-configured (similarly, protection paths can be pre-configured). The result is predictable, resilient behavior from a packet network that is based on Ethernet only and therefore enjoys similar cost advantages,” explains John Hawkins, senior marketing manager for carrier Ethernet at Nortel, the sole vendor behind PBT. The need for spanning tree protocol thus goes away for those engineered paths, and the network becomes more well-behaved and predictable, according to Nortel.
It’s been widely known for some time that BT has been driving the PBT work at Nortel. In fact, BT and Nortel have done joint presentations on PBT in ITU-T and IETF CCAMP. But BT only revealed officially that Nortel will provide Ethernet gear — including PBT technology — for its 21st Century Network in mid-January. The value of the deal, which includes Nortel’s Metro Ethernet Routing Switch 8600 and Metro Ethernet Services Unit 1850, was not disclosed. Hawkins also told xchange there are several trials of PBT under way with service providers around the globe, but he declined to disclose the names of those involved.
Despite any advances PBT has made, it’s still very early in the standards process. However, while the IEEE 802 study group recently began looking at PBT, Hawkins says the technology won’t require much change beyond what already is adopted for 802.1ah (MAC-in-MAC or PBB).
Another View
While the above-mentioned vendors are trying to push forward T-MPLS or PBT, others in the industry — like Atrica Inc. and Juniper Networks Inc. — don’t see the need for either technology. These companies say MPLS and carrier Ethernet solutions available and in use today more than adequately address the requirements of transport networks. And, while PBT and T-MPLS are still in their formative stages, MPLS is already a mature and deployed technology.
Click to Enlarge PBT functions by turning off MAC learning, broadcasting unknown and STP, and using a management plane (or optionally a GMPLS control plane) to populate the switch bridging tables for a specific range of VID/MAC addresses. |
According to Juniper’s Kireeti Kompella, T-MPLS is positioned as a way to make MPLS more palatable to telcos, who are more comfortable with connection-oriented networking. “Just speaking with an IP-centric language scares a lot of people,” he says.
 Atrica’s Umesh Kukreja |
But MPLS can be connection-oriented, using RSVP, or it can be connectionless, via LDP, Kompella adds. The funny thing about the effort to make MPLS more connection-oriented, he continues, is that most service providers using MPLS in their networks are using LDP, so this T-MPLS effort “goes against the grain of what people are actually doing in their networks.”
Saying he doesn’t see T-MPLS gaining any traction in the marketplace, Kompella concludes that “T-MPLS offers nothing significantly different than MPLS except they’ve made a few annoying, subtle changes.”
As for PBT, Kompella says this effort by Nortel to make Ethernet more carrier-class by going from a more connectionless to a more connection-oriented construct is also problematic because there are no features in PBT that aren’t already in MPLS. And, he adds, PBT requires a whole new control plane and data plane, plus it doesn’t include the fast reroute function found in MPLS. “PBT is what MPLS did 10 years ago,” Kompella says. “PBT is trying to redo all kinds of standards like pseudowire, Layer 2 and 3 VPNs, etc. The point is PBT is now where MPLS was 10 years ago.”
Umesh Kukreja, director of marketing at Atrica, says of the two technologies, his company is more aligned with T-MPLS simply because it’s more standard and a better match with what the company already has. But he says both T-MPLS and PBT essentially are doing what Layer 2 MPLS already has defined in terms of traffic engineering and other aspects. He adds that both PBT and T-MPLS have been, to date, positioned and discussed for point-to-point services only.
Nortel's View | |
| Ethernet offers the lowest cost per bit but has several challenges: | Nortel’s Metro Ethernet Solution with PBB and PBT |
| Services scalability and customer segregation/hierarchy/security |
|
| Protection and spanning tree challenges: — Poor convergence/protection/ reliability — Stranded bandwidth |
|
| Hard QoS and traffic engineering |
|
| Service management and OAM |
|
| Services supported |
|
Source: Nortel Networks Ltd. | |
How Things Stack Up | ||
| Category | PBT | T-MPLS |
| Standardization |
|
|
| Technology |
|
|
| Management | Requires sophisticated management system for tunnel creation and bandwidth management (call admission control) | Requires sophisticated management system for tunnel creation and bandwidth management (call admission control) |
| Tunnel-Forwarding Decision | Based on backbone-provider MAC DA and backbone-provider tag | MPLS |
| Service-Forwarding Decision | Based on I-SID | Based on PW/VPLS label at bottom of stack |
| Protection |
|
|
| OAM |
| ITU-T’s Y.1711 (OAM mechanism for MPLS networks)
|
| Comparison to IETF’s MPLS Tunnel Architecture |
|
|
Source: Atrica Inc. | ||
| Links |
| Alcatel www.alcatel.com Atrica Inc. www.atrica.com BT www.bt.com Juniper Networks www.juniper.net Nortel www.nortel.com Tellabs Inc. www.tellabs.com |
