[lug] Redundant net links

[Nate Duehr]

They don't drop /20 or smaller, they aggregate them.  It's a limitation
of RAM available in backbone routers and processor power to keep track
of very large routing tables.

And you can work with your carriers to get it done right, even with
/20's and smaller.

The best thing is to do a LIVE test of your BGP solution with people
trying to get to the sites from dialups and other stuff "out there
somewhere" on the Internet, preferrably on carriers OTHER than the ones
you use for your backbone.

So you actually drop a link (or just tell the router to stop advertising
it) and see if everything keeps coming in (after a few minutes for
reconvergence).

Of course, it's not quite THAT easy either... BGP Dampening will go into
effect after a particular route flaps a few times and then that BGP
route won't work for quite some time after that (carrier dependant on
their configs).

Another common misconception that I'll throw out here is that many
people think BGP will load-balance connections.  Nope.  You may see 99%
utilization of a particular link and nothing on the other side (rarely,
though), especially if all your users come from a large provider like
AOL and stay mostly on a particular backbone. 

Another limitation of BGP is that it doesn't understand degraded routes,
nor does it understand routes with different bandwidth rates.  It picks
by AS# hop count only.  There are times when you can prepend AS# hops
and other tricks to get traffic to go where you want it by default, but
it's a pretty dumb protocol.

Someone needs to convince the big manufacturers to come up with a new
protocol that does everything BGP and OSPF do and also check links for
throughput in some fashion regularly.  Of course, that extra traffic
would have to be paid for by someone...

:)

Or maybe that protocol is out there and I just don't know about it yet.
Time to get myself to another NANOG conference...!  :)

Hope it helps...

On Thu, Aug 31, 2000 at 11:31:35AM -0600, Sean Reifschneider wrote:
> On Thu, Aug 31, 2000 at 10:34:14AM -0600, Kyle Moore wrote:
> >people that new IP. So they would call in and we would tell them to type
> >in this IP. That just isn't acceptable and I know it is not the way to
> 
> As I understand it, for you to be able to do what you want you need:
> 
>    To get an Autonamous System Number (ASN).
>    Get a /20 block of IPs (apparently many large providers drop route
>          advertisements for anything under a /20).
>    Set up BGP4 (which may require you to have a fairly hefty router, I
>          think you'll need something with at least 128MB of RAM).
>    Sign up with both your up-streams for BGP4 service.
> 
> or:
>    
>    Co-locate your box a somone's facility that does the above.
> 
> If you want to do it on the cheap, you can set your DNS propogation settings
> to be rather low, and then if you're having a problem with one line, modify
> DNS to hit the second one.
> 
> Sean
> -- 
>  George Washington was first in war, first in peace -- and first to
>  have his birthday juggled to make a long weekend.
> Sean Reifschneider, Inimitably Superfluous <jafo at tummy.com>
> tummy.com - Linux Consulting since 1995. Qmail, KRUD, Firewalls, Python
> 
> _______________________________________________
> Web Page:  http://lug.boulder.co.us
> Mailing List: http://lists.lug.boulder.co.us/mailman/listinfo/lug

-- 
Nate Duehr <nate at natetech.com>

GPG Key fingerprint = DCAF 2B9D CC9B 96FA 7A6D AAF4 2D61 77C5 7ECE C1D2
Public Key available upon request, or at wwwkeys.pgp.net and others.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 232 bytes
Desc: not available
URL: <http://lists.lug.boulder.co.us/pipermail/lug/attachments/20000901/ad94916a/attachment.pgp>