The Next Generation of Routing Architecture

New to me, LISP Looks Very Promising

Cisco announced a business-oriented pad somewhat similar to IPad. While intriguing, it's not the most intriguing new thing I learned today at networkers, so let me write briefly about LISP. I will be brief today. Tuesday at the show is always busy. The big party for all attendees is Wednesday night, and monday night is free beer and food on the show floor, so all the folks doing smaller get togethers try to schedule Tuesday night.

LISP - not the old programming language, but instead Locator ID Separation Protocol - may well revolutionize routing at the edge of Enterprises, homes, and even individual mobile devices. The first session drew 400 folks. I went to a panel discussion today, which was mostly a Q and A session. I'm still shaking my head in wonder.

Perspective: enterprises that connect to multiple ISPs have several problems that BGP just cannot address well. The enterprise typically gets an address block from an ISP. But which ISP? Once done, a movement away from that ISP requires renumbering the IP address space. Regardless, load balancing inbound traffic towards the Enterprise works poorly, typically meaning that one link is mostly unused during normal ops.

LISP solves both problems. In the demo, it took 8 lines of config in IOS, and that will be smaller soon. Also, LISP provides FIB expansion relief in the Internet core.

The migration path lets you get the benefits of LISP even before anyone else in the Internet implements it. (See Facebook mention below.) LISP is in development by Cisco ans others, and working in the Cisco lab, and likely to be in the published IOS and NXOS code by early 2011. I think you can get early code now to test.

Facebook is already one of the early adopters. Go to, or substitute a 6 instead of for the IPv6 site. That's LISP, done by Facebook. When you connect, routing into this Facebook site uses LISP; the opposite direction's routing logic is unchanged.

There is much more to LISP than what I wrote here. In summary, you can get a small public address range, that you keep forever, and move to any ISP. You get good balanced inbound balancing. You may not need to run BGP at the Enterprise any more. And it is intended to be easy to implement (the demo showed a very simple config).

It is still early for LISP, but it looks like a big deal. As summarized by one of the panel members in the session, it may well be "The next generation in routing architectures." also check out, and


Copyright © 2010 IDG Communications, Inc.

The 10 most powerful companies in enterprise networking 2022