Wednesday, February 20, 2008

Understanding MPLS VPNs

If you are an enterprise user of carrier WAN offerings, it is likely you've been offered MPLS as a solution. Most carriers are encouraging customers to consider MPLS based services over traditional Frame Relay, ATM and even Point-to-Point transport.

One misconception is that consumers of MPLS carrier services must "run MPLS" within their own networks or at least on the edge device(s) connected to the carrier's MPLS service. This is not the case (unless you are doing something pretty unusual). Standard routers -- and even bridges -- are used on the customer-side. The configurations may be a bit different than you're used to but they're still relatively straightforward (and cookie cutter once you do one).

Your layer 1 and layer 2 skills will still come in handy. The underlying transport is still going to be TDM (DS1, OC-3c, etc). You may end up running a dynamic routing protocol (BGP, OSPF) with the carrier's network. If that's new territory, don't worry. This BGP configuration is far less elaborate than that needed to (prudently) bring up BGP for Internet multi-homing.

So even though the MPLS component will be outsourced to your carrier, to be an informed buyer and troubleshooter when shit-hits-the-fan, you'll want to understand the different ways that MPLS can be delivered and used by carriers to provide your service. When the carrier asks you to make some choices or you're evaluating a prospective solution, you're more likely to get what you need (and hopefully less of what you don't).

Jeff Doyle, the author of Routing TCP/IP Volume I and II (both 900+ pages each), has two quick articles about MPLS. In Part I, he covers the basics relevant to any MPLS user as to the different types of MPLS network options and in Part II he covers some of the nitty gritty relevant to service providers and those with an interest in what goes on behind the scenes.

P.S. I have some experience with WANs. Feel free to ask me questions. You can post a comment here or drop me an e-mail.


No comments: