From: Howard C. Berkowitz (hcb@gettcomm.com)
Date: Wed Dec 24 2003 - 20:14:51 GMT-3
At 2:54 PM -0800 12/24/03, Shafi, Shahid wrote:
>Yes CCIE2b,
>
>I am just going through "BGP Design and Implemetation". The book
>approach is Case-Study based and there are lot of configs examples all
>over. I still feel it is a OVERKILL for CCIE Lab though. But no doubt
>its worth the investment if you want hands-on approach to BGP.
>
Eeek. And I consider it UNDERKILL for real world BGP, at least for
any serious ISP applications or even complex enterprise
backbone-of-backbones.
I'm not sure what you mean by "hands-on" in this context. Personally,
I didn't really understand BGP until I backed up and really got
familiar with routing policy, then RIPE-181, now the Routing Policy
Specification Language: http://www.isi.edu/in-notes/rfc2622.txt or
the tutorial http://www.isi.edu/in-notes/rfc2650.txt. Understanding
(and participating in) RPSL at least let me have a real understanding
of what routing policies do, although it took a good deal more work
in operations forums to feel comfortable explaining all the tools
relevant to Internet operations, ranging from justifying and
obtaining IP address space and AS numbers, to tracking IP allocations
such that you can get more when you use it up, to multiprovider
peering and how exchange points work, etc.
As a side note, I recently gave a private class that would have
involved labs on the customer-ISP interface. There were some
unrelated hardware problems that prevented setup, but about the
minimum configuration that I could build of a "simulated Internet"
took at least 6 routers, each with the ability to have a good many
subinterfaces via VLAN or Frame switching. I would really have liked
a Zebra box, probably front-ended with a router, and run additional
services such as a routing registry either on the Zebra box or other
UNIX boxes. You need to have the ability to have at least 5 routers,
each running a different ASN, to show significant AS path issues. At
least one AS should have two or more physical routers to show
multi-POP issues, and obviously even more if you are getting into any
complexity of route reflection.
In other words, at least one or more CCIE pod-equivalents to generate
the external routes. As long as the routers run BGP, they don't have
to be very big if you frame switch them, although old routers might
not run the images with features of interest. My setup was mostly
3640's, but that was what was on hand.
I'm looking into the possibility of virtual classes with such a setup
and curriculum, but haven't yet decided if there's a market.
This archive was generated by hypermail 2.1.4 : Sat Jan 03 2004 - 08:25:44 GMT-3