At the service providers across the globe, the most widespread and native selection is MPLS and not MPLS-TP. The former has broader acceptance, superior interoperability, and operational intimacy for the engineering folks.
In Asia, particularly China, MPLS-TP has taken off in a big way. While some industry observers view that part of the world as exceptionally unpredictable and may suggest that “TP” could even be replaced, it seems that at least one of the large Chinese operators has such a critical mass of deployment that such a move is not imaginable.
On its face, one would think customers should want to go with MPLS-TP over MPLS because the former is a newer version. Ordinarily, any supplier that would happen to start product development from scratch would definitely be inclined to begin in the “TP” direction. Theoretically, MPLS-TP should be a slam-dunk in a greenfield situation at a service provider.
Obviously, there is also the well-known advantage of MPLS-TP in IP network interoperation in which there is a lot of Ethernet over SONET/SDH. There has also been the proposition that there would be a more of a need for the latest solution for over 100GbE situations. (Thankfully, the hype has died down on higher data rates coming to the market in the foreseeable future.) Many of the optical system companies offer MPLS-TP and interoperate with current MPLS networks.
Certainly, a lot of customers that have already installed an MPLS network in the core or the edge would really like to have straight MPLS out to the access. For buyers that are coming at a situation with more of a transport predisposition with centralized control/provisioning, the notion that has been promoted is that MPLS-TP is a viable option.
Somewhat ironically, or perhaps not so paradoxically, although the incumbent operators should prefer MPLS-TP for transport-fixated applications because they act more like familiar connection-oriented protocols (OTN, SONET, etc.), and they extend the MPLS domain without the Layer 3 and higher complexity involving a complete suite of dynamic label switching, the historic inclination not to mess with an established layer (MPLS) in the existing network trumps everything else. Moreover, when are incumbent carriers ever in a hurry to remove legacy gear anyway to achieve the promised OPEX savings, not to mention in a rush to speed up the convergence of services?
Of course, the new regime (the Googles, Facebooks, etc.) would prefer to eliminate any variant of MPLS totally. With NFV/SDN and Openflow aiding in flattening the network, why hesitate to go with no protocol over DWDM with CDC as the solution? In fact, one or more optical equipment vendors definitely have that option in mind in terms of development efforts.
As an aside, it is important not to mix up SDN, which is an abstraction layer with a separation of the data and control planes involving provisioning services and flows at the network level – and MPLS, which is a routing communications as well as an information distribution protocol.
[written by Mark Lutkowitz]