Along the lines of this thread I have something to ponder....
Suppose we have the situation as posed below but I need to manipulate how
data comes into my AS not out of it.
I have two links to two separate ISPs with two separate AS numbers. My
links are not identical (say one is a T1 and the other is a 512K frame
circuit) and I need to make sure that the majority of the traffic coming
into my AS is from the T1. I know there is the MED attribute
(non-transitive) that can influence my neighbor AS to prefer a specific
path from two different paths in the same AS but what about two different
AS's. How can I get them to "share my preference" ? I saw in the Cisco
docs the Always compare MED command but I'm uncertain that applies in this
situation as the MED is non-transitive. I've read up and have been told
that this can be handled with route maps but I'm not sure how that would
affect external AS's.
ISP #1 AS -------------512k--------------------(My
AS)------------------------T1----------------(ISP #2 AS)
If someone has done this or has any insight would you post any sample
configs?
Thanks,
Kris
-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]
Sent: Tuesday, September 05, 2000 8:41 AM
To: [EMAIL PROTECTED]
Cc: [EMAIL PROTECTED]
Subject: Fw: BGP on the Brain - Design Issue
Maybe I can help with this.
BGP doesn't do load balancing in the traditional sense of the word. The
only time you'll see a router have two BGP learned routes to the same
network is when that router has two connections to the same ISP and you
have configured the command "maximum-paths 2" under the BGP routing
process. And I venture to say that you'll never see a BGP router doing
"per-packet" load balancing. Not in the real world. Per-packet load
balancing requires process switching which drags performance of the router
down.
Typically, to make sure that both links get utilized, the edge BGP routers
will advertise a default route into the IGP and from there it is up to the
IGP to make the best routing decision. For instance:
<BGP>------Router1------<OSPF>------Router2------<OSPF>
--------Router3------<OSPF>-------Router4----------<BGP>
Both Router1 and Router4 are injecting 0.0.0.0 into OSPF. Now Router2 will
prefer the 0.0.0.0 coming from Router1 because it has a lower OSPF Metric
than the one coming from Router4. (and the inverse is true of Router3).
Of course, you as the Design Engineer have every option open. You can
influence all the traffic to go one way or another. But you have to know
the network and the traffic patterns very well before you start. You can
set metrics to influence path selection. Or you can use Route Maps (Policy
routing). You can have Router2 Send all HTTP traffic over Router1's
default link, and all other traffic over Router4's link.
You can customize BGP to suit your needs, but the difference between it and
other Routing Protocols is that you have to decide the direction, you can't
leave it up to BGP to pick the best path (like you would with an IGP).
___________________________________
UPDATED Posting Guidelines: http://www.groupstudy.com/list/guide.html
FAQ, list archives, and subscription info: http://www.groupstudy.com
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
