Simulating full BGP peers

Sean Finn seanf at cisco.com
Sat Jul 11 00:57:31 UTC 1998


Oleg, 

    you raise valid concerns; and I recognize that
I jumped in with a recommendation without asking
the appropriate qualifying questions. (i.e.,
"What kind of question are you trying to answer?")

Things that Netsys can do:
  * Collect and/or model observed BGP tables, 
    and their redistribution into IGPs. 
  * Evaluate and validate connectivity.
  * Quantify the size of the effective 
    BGP tables.
  * Detect many BGP configuration errors
  * Provide explanations, and recommendations
    for resolution for some common configuration
    or connectiviy problems

Things that it doesn't do:
 * Model/Estimate router memory requirements 
 * Attempt to quantify transient conditions,
   or their effect on router CPU utilization

I'm personally curious as to exactly what 
kind of question Jon is trying to answer ...

Hope this helps. 

cheers --- Sean

At 07:27 PM 7/10/98 +0400, Oleg Tabarovsky wrote:
>Hmm, I'm not sure that Jon will be satisfied by
>pure modeling. I understand that he is willing
>to inject either random or actual core size
>routing tables into real lab network. Netsys will
>be not of much help in this case. (Say if you want 
>to monitor CPU load figures depending on route flap
>or memory occupied by paths from 3 peers).
>Another issue: IMHO, Netsys is
>not right tool for modeling ISP networks at all.
>If you have arguments to convince me  - I'll be glad to get
>them.

Jon Green originally wrote:
>>> Does anyone know of a way to simulate about three 
>>> BGP peers with full Internet routing tables in a 
>>> lab environment?  Preferably this would be something 
>>> I could run on a Sparc.




More information about the NANOG mailing list