[e2e] TE using IGP metrics

[Olivier Bonaventure]

Hello,

> Thanks a lot. Now, any ideas from the veterans on creating a simulation
> environment for the performance analysis of such schemes. Initially we
> thought about creating a "completely random" network, i.e., a network which
> has a given number of nodes and given number of links but the connections
> between those nodes are completely random. The traffic sources and
> destinations are also random and so are the link metrics. It didn't sound
> like a good idea.
> 
> Any work or suggestions in creating a network where such simulations can be
> run. Purpose of simulations would be evaluating performance of traffic
> engineering algorithms based on IGP metrics. The performance would be in
> terms of recovery time in cases of failures. Are there any papers which
> suggest using some network and traffic models for running IGP simulations?

Classical network models include :

BRITE : http://www.cs.bu.edu/brite/
GT-ITM : http://www.cc.gatech.edu/projects/gtitm/

An alternative is to start from network maps such as those gathered by
the rocketfuel project
http://www.cs.washington.edu/research/networking/rocketfuel/

The second issue is the traffic matrix. I haven't seen any model of
traffic matrix and the only publically available information about
traffic matrixes is the Abilene website :
http://www.itec.oar.net/abilene-netflow/

The papers that I have seen usually start with proprietary traffic
matrix information and perform some scaling or permutations on this
matrix.

Best regards,


Olivier Bonaventure

-- 
CSE Dept. UCL, Belgium - http://www.info.ucl.ac.be/people/OBO/