[c-nsp] PE NAT / VRF Aware NAT on PE

[quinn snyder]

> On Sep 30, 2015, at 02:06, Nick Hilliard <nick at foobar.org> wrote:
> 
> the advantage - as with switch stacking - is with administration.  You end
> up with O(1) admin interfaces to lots of boxes instead of O(n).  This can
> be important if there's too much overhead associated with maintaining
> parallel installations of IOS and XR.
> 
> Personally, I'd be more concerned with the loss of path redundancy and loss
> of service in the case of upgrades.
> 
> Nick

absolutely. 
however -- as with any control-plane sharing feature -- there are also some tradeoffs, namely around upgrades and sometimes erractic behaviour of a chassis because of its peer. 
knowing and understanding the pros and cons of moving to nv is important (although its the same for any feature) -- and not just looking at nv as a panacea is critical for success in the network. 

i tend to be of the mindset of 'separate boxes have separate control-planes' -- so i carry that prejudice in to this conversation. 

i do think that touchpoint minimization is "a good thing"(tm) and welcome the use of automation/provisioning systems that interact with the control-plane of $device, especially if administration is already ornerous within the network. however -- i just tend to be wary of control-plane unification technologies, as i have been bitten and burned by things not working as expected due to said technology. 

q. 

--
quinn snyder | snyderq at gmail.com

-= sent via ipad. please excuse brevity, spelling, and grammar =-