Inter-DC IP-based vMotion with LISP

In early autumn of 2010, a “DRAFT on Cisco Nexus 1000V LISP Configuration Guide” appeared on CCO. It’s gone now (and unfortunately I haven’t saved a copy), but the possibilities made me really excited – with LISP in Nexus 1000V, we could do close-to-perfect vMotion over any IP infrastructure (including inter-DC vMotion that requires stretched VLANs and L2 DCI today). Here’s what I had to say on this topic during my Data Center Interconnect webinar (buy a recording).


  1. Has anyone networking vendor other than Cisco showed any interest in using LISP? Not saying I think it's a good or a bad idea. Personally I think it has some cool features.

    My question is more to do with they fact that while yes drafts have been submitted to the IETF, if you read the drafts, or even go to the meetings. It's just Cisco and Cisco customers.

    I'm worried that with the emotional backlash against Cisco, that no other vendors will ever touch it.

    No matter how cool it is, how much help is it going to be if ends up being like EIGRP
  2. Let's put it this way: either the industry adopts LISP or we'll have some "interesting" problems a few years from now when IPv6 multihoming explodes BGP routing tables (not to mention provider edge router TCAMs). I don't see a viable alternative anywhere near the production-grade state at the moment.

    As for "industry adoption" - if Juniper decides to jump on board, we have all we need. LISP is primarily a technology solving SP problems, DC use case is more like "gee, I have a hammer and this sure looks like a nail" approach.
  3. Oh, I totally agree, it's a nasty problem that needs to be solved. Well..unless you make memory chips, then this is a good problem ;-)

    And I think the LISP team deserves a lot of credit for taking a very methodical approach to testing, sharing, and involving people in LISP.

    But you exactly hit my point. Juniper seems to me ( just based on working there for a year a few years ago) to be much more likely to come up with an alternative than adopt something so closely associated with Cisco.

    I think if LISP had be done by a different group, or maybe had been started out of universities like OpenFlow, it would be seeing HUGE support.

    I just fear that the "anything but cisco" mindset is going to keep even good things from geting the acceptance they probably deserve.
  4. Jon,
    Your "fear" sounds a little more contrived than genuine. Just say'n.
    As Ivan said, if Juniper feels LISP is missing the mark, they can certainly initiate a new standards effort as well. But that hasn't happened yet.
  5. While I am happy to admit I am not a big cisco fan. (I was a cisco customer from 1996-2004, generates strong feelings) in this case I really am worried.

    In the edge router market cisco has about 37% market share. As for Core router market in 1Q FY11 numbers reported are

    Cisco #1 with 7% growth

    Juniper Networks # 2 with 13% growth

    Huawei at #3 with 34% growth

    So lets be more positive, since these are new sales and not existing installs and say cisco has 50% market share.

    Now, do you see what I mean? If only 50% of the routers on the net, 37% of new routers being deployed are cisco and can use LISP. And if Juniper, Huawei and others (others are less than 3% of the market) do not use or ever support LISP, then 50-73% of routers will not be running LISP.

    So then how much can LISP really help? LISP is a good idea, they did a good job of getting it out there, testing it, involving customers.

    I'm not worried about LISP not working, or not being good. I'm worried that a good idea may not get a chance to shine not because of the idea, but because of who delivered the idea.
  6. I think there are two separate value propositions here with LISP. One is for the end user and the other is for the Internet at large.

    For an end user, as long as there is one other router running LISP to communicate with, they can see immediate value. I know this because we operate LISP gateways specifically for this purpose.

    Running LISP this way, however, is transparent to the Internet at large, and it doesn't benefit from LISP until the separate islands of LISP federate using a DNS-like solution called the DDT.

    I actually presented this at NANOG55 if you are interested in the details.

  7. I think it would be cool if Cisco can add LISP support to the Nexus 1000v but unfortunately that has not happened until today. I think the next best thing is running LISP on the CSR 1000v which can be used as a default gateway for the VMs.

    For those interested I have written about a use case for LISP on CSR 1000v showing how to migrate a VM from a data center to the cloud without changing its IP address
Add comment