How viable is it to take IP out of IP/MPLS?

Mark Seery, VP, Switching and Routing

It is technically possible to take IP out of IP/MPLS, but the cultural and political problems are much greater.

The recent Ovum RHK brief Economic forces splintering the IP/MPLS convergence thesis (February 2008) raised the potential of equipment vendors considering the possibility of driving MPLS to be the lowest-cost technology by taking IP out of it. This raises a question about the viability of this approach.

There are at least three important dimensions to this question:

• technical
• standards
• cultural.

Technical

If an infrastructure element does not intend to exchange routing information with other networks, including customer routers, IP is not a requirement from a technical perspective. IP is not needed to process control plane messages with other infrastructure elements, though it might be convenient, and IP does not need to be processed in the fastpath/user/traffic plane other than to pack it into some other transport protocol/technology. This has been established time and time again with technologies like SONET/SDH, ATM, and Ethernet. So, there is no technical issue if the focus is transparent transport.

Standards

The recent fracture in the standards community over T-MPLS (an ITU modification to IP/MPLS) indicates that any attempt to even modify IP/MPLS outside of the ITU, let alone the ultimate heresy of taking IP out of IP/MPLS (which I believe was the T-MPLS endgame) is a huge problem. My understanding is that if T-MPLS work continues within the ITU, all relations between the IETF and the ITU on any subject may be at risk.

From many people's perspective, T-MPLS is essentially dead at this point (outside of IP/MPLS profiles being created with the IETF that may eventually be coined with a similar name).

So, developing any standard that is not essentially the same as IP/MPLS cannot be called something-MPLS without creating a schism in the standards community. So much is in the naming (and one of the main drivers of T-MPLS told me months ago that he did actually want to call it something else).

Cultural

If equipment vendors go it alone, calling the standard something other than MPLS, but marketing it as being MPLS-like, that will raise cultural tensions within companies doing it, and within the industry as a whole. For some this seems like turning the clock backwards, or a return to the days of ATM.

Summary

From a technical perspective, removing IP from IP/MPLS is doable (from a technical perspective, almost anything is doable). From a standards perspective, it is very difficult and probably impossible. From a cultural perspective, it is very difficult.

When speaking privately, I have often said T-MPLS was a non-starter - not because of technical issues, but because of standards/cultural/political issues. I believe events have demonstrated that to be the case.

It is possible this is simply an option the industry will never pursue. If that is the case, the industry must return to the question of how inexpensive IP/MPLS can be made (and how inexpensive suppliers want to make it) versus Ethernet architectures.

The year 2008 is shaping up to be an interesting one from this perspective.

Mark Seery has 25 years of industry experience, covering both private and public networks in the areas of network operations, software development, product management, and product marketing. Mark's current work focuses on technology, products, and network architecture in the areas of IP/MPLS, Ethernet, and ATM. Mark also works with teams doing research in the areas of optical transport (SONET/SDH and DWDM) and access. In addition, Mark regularly comments on a wide range of industry trends including industry structure and evolution.