Is LISP Going to Save the Internet?

A couple months ago, after attending the FutureNet conference, I wrote a blog about the impending Internet meltdown. In short, there are two problems afflicting the Internet:

  1. We are running out of IPv4 space (we knew that).
  2. The global Internet routing table is too big now and getting bigger fast.

Either of these could lead to Internet outages, brownouts, and changes in expected access in the coming years. So, smart people around the world are working on fixes. One of them, developed by Cisco engineers, is LISP - the Locator/ID Separation Protocol. I've heard about LISP in the past, but never took the time to study it until a nice article in Cisco's IP Protocol Journal sparked my interest. The article was good, but I needed more. In particular, the article didn't explain how LISP was going to fix the two biggest problems the Internet faces, which I knew LISP was created to fix. So, I found a very good, albeit long video, of a tech talk with one of the creators of LISP, Cisco Fellow Dino Farinacci:

LISP is a map-and-encap protocol that separates the location from the endpoint ID in an IP Address. For example, currently, an IP address - let's say 80.248.26.3/24 - identifies both the location and the endpoint. It's on the 80.248.26.0/24 network and it's the .3 host. The problem with this is that to tell other devices around the Internet the location of this host one must send a network advertisement (80.248.26.0/24) into the Internet core routing table. If you're a good Internet citizen, you get a nice big block of address space from your ISP and you aggregate 80.248.26.0/24 into 80.248.16.0/20. The ISP then aggregates into 80.248.0.0/16 and the core routing tables are nice and efficient. But let's be honest....it's not that you don't want to be a good Internet citizen, you just can't. You really only have 80.248.26.0/24, not the entire /20. So, instead of a nice /20, you advertise a /24 into the Internet. Plus, you didn't want to be tied to an ISP forever by renting their address space, so you went and got 80.248.26.0/24 on your own from ARIN. And, finally, no good WAN design is done with one carrier. So the site with 80.248.26.0/24 in multi-homed with connections to two ISPs, both advertising 80.248.26.0/24 into the Internet core. Congratulations, you have just contributed to the downfall of the Internet. It's not your fault. You're doing what's best for your organization, not the Internet itself. LISP attacks this problem by separating the location from the endpoint ID. Let's say this is your packet: Source = 216.215.54.50 Destination = 80.248.26.3 This packet flows normally in your organization, probably following a default route, to your Internet routers. With LISP, once the packet reaches the Internet routers, your packet is now mapped-and-encaped. First, your Internet router, now dual functioning as a LISP Ingress Tunnel Router (ITR), looks up a single IP address that says it can reach 80.248.26.0/24. That IP is 20.20.20.20. Your Internet router (the LISP ITR) then encaps your packet setting the destination to 20.20.20.20. LISP Source = Your Internet Router LISP Destination = 20.20.20.20 Source = 216.215.54.50 Destination = 80.248.26.3 Now, the only routes that need to exist in the Internet core are a few thousand host routes - in this case 20.20.20.20 - that identify LISP routers that know how to get to endpoints. You can read the article to find out how LISP builds its "map" table. If you're thinking LISP is a tunneling protocol, you're right. No, you won't see something like this in your router any time soon:

interface tunnel0
 tunnel protocol lisp
 ip address x.x.x.x.

But LISP does wrap your packet in an outer shell to get it across the Internet.


But now back to the point of LISP: will it save the Internet from problem #1 and #2 above. Umm, I don't see it. I need to do more reading on LISP, but it definitely doesn't solve #1. Honestly, only IPv6 will do that and no one is jumping on the bandwagon yet. As for #2, it could happen, but will take a lot of coordination and buy-in. What's the incentive for a single organization - like the one I described above that is currently breaking the Internet - to move to LISP? Carriers could demand that all packets flowing over the Internet be encaped in LISP, but that's not likely, not when carriers are trying to make money by getting new and keeping existing customers (hint: you don't tell customers what to do if you want to keep them). Government could demand it, but my libertarian side shivers at the idea of 100 Senators trying to get that right, let alone the total unconstitutionality of it all. Plus, even if the US government ordered it, what's going to make China, India, the EU, and the rest play along? So, I'll keep reading, but for now, I don't see it.

More >From the Field blog entries:

Cisco Efficiency Assurance Program

CCDE Practical is a Go!

Brocade Buying Foundry...Has to be an FCoE Play

I Need Some Twitter Friends

Wireless Steps Up

  Go to Cisco Subnet for more Cisco news, blogs, discussion forums, security alerts, book giveaways, and more.

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.
Related:
Take IDG’s 2020 IT Salary Survey: You’ll provide important data and have a chance to win $500.