[c-nsp] IP SLA?

[Dan Brisson]

On 3/24/15 8:30 AM, Jared Mauch wrote:
>> >On Mar 24, 2015, at 8:27 AM, Dan Brisson<dbrisson at gmail.com>  wrote:
>> >
>> >I'm curious what folks do in the situation where you have redundant links to your customers.  I'm speaking primarily in co-lo environments where you offer redundant Internet connectivity to co-lo customers.  So for example, you give a customer 2 ethernet handoffs from two separate Layer 2 switches.   Now what do you do if the customer wants to go to a routed model using both links.  I could allocate /30s for both links, but then I have the issue of how to reliably route their block to them w/out running a routing protocol that will detect if one of the links goes down.  That's where I came to static routes with IP SLA but I wanted to make sure I wasn't missing something easier.
> Do they have two routers as well, or a simpler subnet config?  Perhaps something like VRRP and using a protocol to inject these ‘connected’ routes to the rest of your network?
>
> - jared
At this point I'm not sure if they will have one or two, although your 
suggestion of VRRP had crossed my mind.  In that case, we both really 
need to run VRRP on each side to protect against failure, right?  I was 
trying to get around allocating a /29 for a VRRP subnet, but I suppose 
one /29 = two /30s.  Just seems like overkill, but I guess that's why I 
asked the question.

It also occurs to me that this really isn't any different than a 
customer buying 2 connections from their premises to the same ISP. Or is it?

Thanks!
-dan