Re: [Idr] draft-jakma-mrai-dep-00 (fwd)

On Tue, Nov 25, 2008 at 10:07:33PM -0800, Pedro Marques wrote:
> The typical example being an SP that acquires a newly preferred peer route 
> which them causes it to send withdrawals to other peers (reachable converts 
> to unreachable).
>
> My understanding is that this behavior has been well document in practical 
> observations of BGP convergence.

That was the example I needed.

Part of the reason why I suggested separate timers is that the above
scenario is within the scope of iBGP convergence.  This means that
within the expcted iBGP MRAI you should be able to safely send
unreachables quicker to your eBGP peers than you'd otherwise be able to
if you were waiting on the (typically longer) eBGP MRAI.

I'd still recommend different timers but I think I've just been
convinced that the withdraw timer should have a lower bound of the iBGP
MRAI value.

> In general my recommendation for MRAI values would be 0 for eBGP and 0 for 
> iBGP. As the draft correctly explains calculating a value that best fits 
> the network is a very tricky business.

And, as long as you're not worrying about route flap dampening, TCP
queue lengths and processing delays is often enough to squelch a lot of
the updates in networks that are "busy enough".

-- Jeff
_______________________________________________
Idr mailing list
Idr@ietf.org
https://www.ietf.org/mailman/listinfo/idr