In my last article I discussed labeled transport, or carrier-of-carrier MPLS solutions. If you recall, labeled transport service is where a carrier will provide backbone services from another carrier via IP. The carrier provisions label-switched paths (LSPs) through its backbone to transport another carrier's IP and IP-VPN traffic. This allows carriers to provide backbone services without provisioning circuits dedicated to the carrier's customers. In this article, I will explain how all this works from an MPLS perspective.
The basic premise of the solution is that the carriers exchange labeled traffic. As discussed in the last article, the carrier-of-carrier solution is ideal for ISPs who want to interconnect geographically distributed points of presence (PoPs) or for carriers who are providing MPLS VoIP services and want to interconnect geographically distributed MPLS VPN service areas.
Traditionally, these customers had to purchase circuits or build out the infrastructure themselves. The cost of these options in some cases have prevented this expansion. Now these customers can connect to a carrier who offers labeled transport and exchange labeled traffic that encapsulates the customer's traffic. In essence, a large virtual circuit is used to connect one carrier's customers over another carrier's backbone.
There are two key parts to understanding how this works. The first is the backbone carrier and the second is the customer carrier. The backbone carrier is the carrier who is providing the backbone service, and the customer carrier is the carrier who is utilizing the backbone services. Let's assume that the customer carrier is an ISP. The customer carrier wants to connect PoPs in California to PoPs in Atlanta. Traditionally this would have required that the customer carrier purchase long-haul circuits between the routers in California and Atlanta. If redundancy is required, that would have meant two long-haul circuits.
As I stated before, an alternative option would be to build out the backbone between Atlanta and California. Both of these are expensive. With the carrier-of-carrier solution, the customer carrier can purchase access circuits to the backbone carrier in both California and Atlanta. The backbone carrier will build a virtual circuit that will encapsulate the ISP traffic inside a labeled packet. It will transmit this across the virtual circuit and then hand it off to the Atlanta ISP PoP.
In this scenario, the California and Atlanta PoPs have CE routers connected to the backbone carrier PE routers. For the backbone carrier PE routers, the CE routers look just like any other CE router. The CE routers exchange routing information over the backbone carrier just like any other MPLS CE-PE routing exchange. These routes are considered internal routes.
However, since the endpoint CE routers are ISP routers, their routing tables could conceivably hold the entire routing table. In order to facilitate the exchange of the entire routing table, the CE routers establish an IBGP session over the backbone carrier's network. These routes are considered external routes. In both cases the internal and external routes are IP. This solution allows the Atlanta CE and the California CE to exchange full Internet routing tables without the backbone carrier routers having to know about it. This is a very elegant solution.
The second scenario is when the customer carrier is a MPLS VPN provider. In this scenario the external routes are VPN-IPv4 routes (not IP). In this scenario, all the routers in the customer carrier network must use MPLS. In the previous example only the CE routers were required to use MPLS. In this case there is a mesh of IBGP tunnels built between the PE routers in Atlanta and California as well as the CE and PE edge routers between the carriers. This allows label exchange from one edge to the other for carrying the VPN-IPv4 routes. This is true label exchange for the customer carrier CE routers that belong to their VPN customers.
This technology is in place today and is being utilized on a limited basis. As discussed several articles ago, the QoS and SLA adherence across multiple clouds is a severe limiting factor. In the next article I am going to discuss MPLS certifications and begin a series of articles that will discuss the MPLS foundations required for certification.
Robbie Harrell (CCIE#3873) is the National Practice Lead for Advanced Infrastructure Solutions for SBC Communications. He has over 10 years of experience providing strategic, business, and technical consulting services to clients. Robbie resides in Atlanta, and is a graduate of Clemson University. His background includes positions as a Principal Architect at International Network Services, Lucent, Frontway and Callisma.
