[c-nsp] non-BGP ISP redundancy

[Rupert Fiennes]

There is one other possibility if you have a hub site to use; run GRE/IPSEC tunnels with
source addresses on the uplinks to the two ISP's from your edge routers to the hub site,
then run a routing protocol over these tunnels. When one link goes down, even if the
interface is "up", the routing adjacency will fail and traffic will failover to the other
tunnel.

Rupert

Bruce Pinsky wrote:
> Grant P. Moerschel wrote:
> 
>>>Greetings,
>>>I have an environment with a full T1 connected to a 1700 to ISP1 and a
>>>frac T1 (256K) connected to a 1700 to ISP2.  I don't have the resources or
>>>address space to use BGP.  But I would like to have some egress load
>>>balancing and some ingress redundancy. I host my mail at this site and my
>>>web servers elsewhere.  I can accomplish smtp redundancy with MX records.
>>>
>>>Any ideas on some good approaches? Thanks
>>>
> 
> 
> Rodney Dunn posted this in another thread:
> 
> "You must have some form of detection to know that one
> of the ISP's is not reachable.
> 
> The best solution to that is BGP or either
> IP SLA with object tracking.
> 
> You could have two default routes in the 2801 each
> pointing to the different ISP's.
> That would give you CEF loadsharing by default.
> 
> However, your return traffic would be a problem
> because depending on what your source address is going
> out that is what your return path would be.
> 
> Most people get around this by doing PAT on the ISP
> interface address so that makes sure the return traffic
> for that flow comes back via that same link."
> 
> --
> =========
> bep
> 
_______________________________________________
cisco-nsp mailing list  cisco-nsp at puck.nether.net
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/