Older: 
The Future of the Internet Core
In the last few weeks there has been a large volume of dialog around which technology should define the “core packet transport” of the Internet. Mostly that dialog has been focused on the ongoing debate between the MPLS technology camp and the Carrier Ethernet camp. If you are not familiar with this debate I have included (at the end of this post) some links from various trade and other web sites that show the level of passion and, in some cases, hostility in this dialog.
The reason I point this out is that one easy way to determine the significance of a technical inflection is the level of defensiveness that emerges when a technology is challenged by a viable alternative. I remember when the Token ring camp (which I was involved in) determined that “this Ethernet thing” was not robust, predictable or mainstream enough to be relevant. I also remember when the Novell IPX and DECnet camps argued that Internet Protocol (TCP/IP) was also not sufficiently mature to be used in the corporate world. Obviously they were both wrong and the world moved forward. Today, most new IT people don’t even know what Token ring or IPX or DECnet are (not to mention APPN, LAT, Banyan Vines, AppleTalk, and others).
What is clear is that the technology that ultimately won each debate and became the common model was the technology that offered the lowest cost, the simplest operating model and the greatest scalability and flexibility to move forward. The industry has always gravitated to technology with those benefits.
Once again the industry is in this debate. We are debating if the future is about the multi-protocol label switching (MPLS) path or if it will ultimately evolve to a model that is Carrier Ethernet over a high-performance optical layer. At its very core the debate is around whether or not we should use Ethernet to transport Ethernet services or if Ethernet is best transported over another technology, such as IP/MPLS.
Today, the industry is split. There are clearly many MPLS carrier customers, but there is also a rapidly growing number of operators using Carrier Ethernet (more than 40 using Nortel’s gear alone). Additionally, there are millions of enterprise customers using Ethernet as their primary transport technology. The contentious point is that Ethernet as defined by IEEE 802.3 and 802.1 was not initially designed for carrier applications. Today, however, the IEEE 802.1ah (Provider Backbone Bridging), 802.1Qay (Provider Backbone Transport) and 802.1ag (connectivity and fault management) amendments in the standards track are emerging to add the needed carrier services to an Ethernet switched network. With these technologies it is very easy to provide most Ethernet VPN services over large-scale carrier infrastructure at much lower capital cost and a simplified operating model.
When we look at the MPLS side we see that there is significant active work and a host of draft documents and requests for comments (RFCs), which are the IETF code for a defined and generally accepted specification or document (note that I am simplifying dramatically the IETF standards process, as defined by RFC 2026, as it is easily one of the most rigorous and complex in the industry and, as such, results in very few fully standardized technologies). Even though MPLS has been in existence in some flavor since roughly 1996 (I was at many of the first meetings back before MPLS as a term existed), it is not fully standardized in all dimensions because of both the complexity and the process to solidify its real objectives.
Additionally, the MPLS model has evolved in terms of what it was originally trying to do. Initially it was designed to accelerate the forwarding of software-based routers. That became somewhat irrelevant as hardware-based routing emerged. The technology then became a mechanism to provide QoS using complex signaling and label stacking. That was of use but largely not used because the industry simply created more capacity and was comfortable with CoS. It then became a technology focused on providing virtualization of subscribers. This is of use but not unique to MPLS; many VPN solutions exist that can do the same thing.
What makes MPLS interesting is that it has a broad set of optional capabilities because it is a conglomeration of many iterative technology streams applied to recreate the TDM carrier network over a packet switched network. It gets even more complex when you look at the actual core of MPLS and realize that there are actually many different approaches under this umbrella, including VPLS, T-MPLS, and a host of other vendor-specific approaches.
The bottom line is that the MPLS world is a complex space. The MPLS working group (to quote the group’s charter) is “responsible for standardizing a base technology for using label switching and for the implementation of label-switched paths over various packet based link-level technologies, such as Packet-over-Sonet, Frame Relay, ATM, and LAN technologies (e.g. all forms of Ethernet, Token Ring, etc.). This includes procedures and protocols for the distribution of labels between routers and encapsulation.”
So, why would there be a huge debate over Carrier Ethernet versus MPLS? Three reasons. One, is that the telecom core is still in an evolutionary phase where there has been broad adoption of the various MPLS technologies. Two, most MPLS networks work reasonably well at providing the services they were chosen for over the past decade. And, three, the carrier landscape is usually fairly slow to adopt new technologies because the capital and operating costs of change are nothing to trivialize when one considers the scale of a national or international carrier’s network. Compounding this is the fact that many equipment vendors have made significant investment in both time and dollars to build up MPLS product lines and the idea that this technology investment may not be suitable for one of the most rapidly growing elements of the equipment market (metro networking) is unimaginable.
The bottom line is that there are as many near-term reasons to defend the status quo as there are longer-term reasons to move to an Ethernet over Ethernet technology, such as Carrier Ethernet.
The purpose of this post is not to educate you on MPLS or Carrier Ethernet, but rather to put into context the dialog, which is now front page news every time a customer makes a decision about one of them and every time a product or standard is launched.
Over the next few months, I will try to update the story as it unfolds and add additional context, but for now I ask you to keep in mind that this debate is really no different than many technology inflections of the past. This, too, will unfold with much debate and passion and, like every technology evolution before, we will see a continued movement to simpler, more effective transport and IT systems. The challenge will be in managing the transition and evolution to what is the inevitable.
References:
IETF MPLS Working Group Link
IETF Standards Process:
IEEE 802.1 Web Site
Face Off: MPLS vs. Carrier Ethernet
Ethernet Goes Carrier ClassWhat’s
Next for PBT-less BT?
Analyst: PBT is Not Dead Yet
Nortel Wins PBB Deal with Verizon
There are half hearted organizational “discussions” about to spin up over switched ethernet vs. MPLS in most carrier organizations, but no one is really wanting a repeat of the past conflicts anytime soon, plus they need to fully amortize all of theis MPLS gear they just got done deploying. 
[…] providers. And so on. Take for instance a blog post by Nortel’s CTO, John Roese, discussing The Future of the Internet Core. Essentially he is hyping PBT by pitting it against MPLS. But I don’t buy it. This […]
June 17th, 2008 at 7:20 pm from Wagging the Dog: MPLS vs Carrier Ethernet