Comprehensive example for LISP (Locator/ID Separation Protocol)

[CRISTIAN LACATUS]

Hello

Does anybody know where I can find a comprehensive example for LISP? The main pages on this topic from cisco.com (http://cisco.com/go/lips, and http://lisp.cisco.com/index.html) are full of marketing materials on how LISP will solve all the routing problems, and references to standards.

The only example I could find has four routers (http://www.cisco.com/en/US/docs/ios/lisp/app/note/LISP_lab_test_app_note.pdf), I am pretty sure there is more to LISP than this very basic example.

Why I am asking – we are preparing for a datacenter move,  and the business requirements call for  servers are to be moved in the new location one by one, with the original IP address. For a few month will will have parts of the same networks residing in two different locations. Layer 2 between sites does not solve the “traffic trombone” problems and can be expensive (OTV on Nexus), a layer 3 based solution looks cleaner.

I know, LISP is a fairly new technology in Cisco routers. Are there any good example available? Would you consider LISP for such a job, or (at least I think so) the implementation is still experimental? As usual, the Cisco sales guy told my executives that Cisco solved all these problems with LISP, and it's trivial to deploy this technology. I have the feeling that is not exactly true...

Thanks,

Cristian

[gschudel]

Hi Cristian

First - thanks for the comments/questions about LISP.

Hope i can help point you in the right direction...

** If you go back to the site lisp.cisco.com (v4 or v6) and click on "downloads" you'll find

detailed configuration examples for several (five) "enterprise edge" reference use case.

These are much more detailed and tailored to real deployments than the Lab Guide was.

(you can find the pdf here directly if you like: http://lisp.cisco.com/LISP-cfg_gde_IOS_1514XB5.pdf )

(The lab guide was developed when LISP was first released in IOS as a 'quick and simple'

way to get LISP fully operational in a lab setting so people could get "hand-on" experience

without minimal investment in resources.)

These reference configurations do not cover the VM-Mobility case however.

** With regards to VM-Mobility, it is true that LISP provides the mechanisms to

handle Layer 3 moves and allowing the host to keep it's IP address. The host

can land on (a) the same subnet (extended across DCs using OTV), or (b) to a new

subnet (and still keep it's original IP - even though it's not 'native' to the new subnet).

The "reference configuration guide" for these use-cases has just been completed this

week and should publish shortly. (check back on the 'download' page).

We've tested VM-mobility extensively, by the way, and have ~20 trials underway.

Also -- and importantly most likely for you -- LISP VM-mobility functionality is _only_

available in Nexus code right now. You've mentioned that this might be an expensive

option (N7K).  VM-Mobility for LISP is now on the IOS roadmap - but i don't have a

time frame for delivery yet.

For you, and anyone interested in more details on LISP or to get answers to questions

you might have, the best option is to send an email to lisp-support@cisco.com and

you'll gain access to the team.

Cheers

gregg

(LISP Team)

[chrisgray1]

Hi Gregg,

As you have tested VM mobility extensively, so you sound like just the guy to ask.

I have put together a design for our DCs, there are 4 zones which will utilise LISP, the main premise is that it provides VM mobility across a pair of DCs in each zone with extended L2 subnets.

So there is currently a site to site L2 extension over direct fibre links and is deemed as a potential point of failure.

The primary use of LISP is to provide reachability over the GWAN if there is a failure scenario and my question relates the multi-cast map-notify-group which is used to signal VM moves.

In my design the N7K on each site is the xTR and MS/MR, they use the map-notify-group to notify each other across the extended OTV link about VM moves.

However I would like to know, If there is a problem with the L2 link and multicast can no longer be received, then the N7K should still have an up to date LISP table in memory and by my reckoning as long as it can still reach the RLOC of the other site N7K over the GWAN, the VMs on the other site should still be reachable?

Could you please hopefully confirm this, or let me know if I've made a mistake over this?

thanks

Chris