Comments inline.
--
Jason Roysdon, CCNP+Security/CCDP, MCSE, CNA, Network+, A+
List email: [EMAIL PROTECTED]
Homepage: http://jason.artoo.net/
Cisco resources: http://r2cisco.artoo.net/
""Howard C. Berkowitz"" <[EMAIL PROTECTED]> wrote in message
news:p05001903b6c6a62de3ce@[63.216.127.100]...
> >This may take some explaining, so just bear with me. Also keep in mind
that
> >I've been having hardly any time to read Halabi's BGP book, but will
> >probably go read some after this.
> >
> >We're connected to a few ISPs which allow BGP peering (which I should be
> >settings up sometime soon).
>
> I need to get clarification here. There are two meanings of the term
> "peer" in the BGP context, the first being a simple establishment of
> any BGP relationships, and the second being an economic relationship
> of equals, where you exchange customer routes without financial
> compensation. The alternative to the second is to buy transit from
> an upstream ISP.
Peering for cost with the larger ISPs who will transit our traffic.
> >We've also got two ISPs that will not peer, nor
> >exchange customer routes with us. One is a free 1.5mbit SDSL connection
as
> >we're one of their VARs, the other is a lame T1 that we're stuck in a 3
year
> >contract for a bit more.
> >
> >Right now, I just use static routes to send traffic out the SDSL
connection
> >as the provider only has a single class B. For the T1 to a much larger
> >provider with address space all over, it's just not worth it to try and
do
> >much with it...
>
> I don't understand what you mean by address space all over, or not
> being worth it. Sometimes the whole motivation for BGP is to
> exchange very specific and extensive address information with
> adjacent AS, at the same time avoiding leaking large numbers of
> irrelevant routes into the global routing system.
>
Well, a large amount of </16 netblocks that don't follow an easy pattern
(such as the single class B that the SDSL provider has).
> >
> >Anyway, here is the thought: I happen to know the admins a at number of
> >other ISPs that are connected to the T1 and some other sites that have
SDSL
> >access to the same provider as us.
> >
> >The catch is that of course we could set static routes out to these ISPs,
> >but it's somewhat risky, especially with the SDSL as even though the
> >ethernet interface it's connected to may still be up, the SDSL line
itself,
> >or perhaps something along the SDSL provider's backbone might be down
> >between us and another of the SDSL customers, but the static routes to
the
> >SDSL link would stay up as the interface is still up. Same is true with
the
> >T1.
>
> I don't think there's any way you can know there is a reachability
> failure in a non-directly-connected link without running a routing
> protocol. In the case of an ISP, that pretty well has to be BGP.
Ok, but how would BGP know which interface to go out? These equal peers
would be connected to my non-default interface.
> >
> >Since neither of these ISPs will peer with us, could we still establish
some
> >routing protocol with the smaller ISPs like us that are connected off of
> >them and want to transit traffic through these lesser used links.
>
> Why not BGP to the smaller ISPs? There might be a need to coordinate
> private AS numbers.
Not opposed to it, just looking for the best solution, and this is probably
how we'll go to start with.
> Remember that the BGP tunnels can be between loopback interfaces, so
> as long as you can reach the loopback in the other AS, and
> appropriately set ebgp multihop, you should be able to run a session
> without the intervening ISP being aware of it. Can't promise what
> the performance would be.
Right, but here's what I'm missing: BGP has to know how to reach the other
loopback to exchange routes with the other neighbor. I'll be able to
exchange their netblocks, and then can I tag those netblocks and use a route
map or something to have the router send it out a non-default interface? I
mean, how will the router know what interface to go out? I'm guessing I'd
set a static route out the non-default interface that will get me
reachability to the BGP neighbor. As this neighbor is out a certain
interface, will the router know to send traffic for BGP routes learned from
that neighbor out that same interface?
> >Otherwise, the netblocks we have would route traffic back through the
ISPs
> >they belong to or that we're announcing them with BGP on.
> >
> >The biggest thing is that it needs to be dynamic. If the route over the
> >common single upstream ISP is down, but the connections to these ISPs are
> >up, routes out to our defaults/BGP peers might still get us connected.
> >
> >Thoughts? Comments? Am I just nuts?
> >
> >--
> >Jason Roysdon, CCNP+Security/CCDP, MCSE, CNA, Network+, A+
> >List email: [EMAIL PROTECTED]
> >Homepage: http://jason.artoo.net/
> >Cisco resources: http://r2cisco.artoo.net/
_________________________________
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
