Modelling a large ISP network with C-BGP

Dear All,

C-BGP is an efficient and open-source simulation tool that allows to simulate the behavior of the intra- and interdomain routing protocols in large ISP networks. C-BGP is able to simulate networks with thousands of routers. A key feature ofco C-BGP compared to other simulators is that it is able to support the complex routing policies that are used by ISPs.

Thanks to the support of France Telecom R&D, we have been able to use C-BGP to reproduce the behavior of BGP in a large Tier-1 ISP network. For this analysis, we have developped several tools to automatically convert the BGP configurations of Juniper (JUNOS) and Cisco (IOS) routers, the IGP topology and the BGP routes in the C-BGP format. With these tools, any network operator can easily build a C-BGP model of his/her network. With such a model, it is possible to perform different types of analysis on the ISP network, such as :

  - predicting the flow of the traffic through the network or determining the traffic matrix based on Netflow data (see article in January issue of ACM SIGCOMM Computer Communication Review)
  - predicting the impact of adding a new peering link on the BGP routes selected by the routers (see article in Nov/Dec issue of IEEE Network Magazine)
  - predicting the impact of link or router failures on the ISP network

You can obtain C-BGP, the conversion tools and several papers describing the utilization of C-BGP to model an ISP network or compute traffic matrices from :

The website also contains several examples, such as a complete C-BGP model of the Abilene Network based on their publically available JUNOS configurations files.

We believe that C-BGP could be very useful for ISPs willing to optimise the distribution and the selection of the BGP routes in their network. Comments, suggestions and questions from network operators are more than welcome.

Best regards,

Bruno Quoitin, Sebastien Tandel and Olivier Bonaventure

I'm I alone to find this a bit spammy?

Olivier Bonaventure wrote:

I'm I alone to find this a bit spammy?

announcement of a free, open-source, and looking very useful tool
for operators looks like one of the most important messages i have
read on these lists this week.



    Didnt say it was... it just that it came off the blue... (its started in 2003).

    Snapshot dont compile yet on Freebsd.

    But those do: libgds-1.1.8-rc2.tar.gz <> (May 17th, 2005), cbgp-1.1.20-rc2.tar.gz <> (May 17th, 2005)

I did spend a hour with the soft:

    . it is all the quality of good software...

    . small binaries...

    . fast...

    . large memory usage for simulation... (~128M for "valid-topology (July 15th, 2004)")

Where I could use it:

    Well, I can see it being used in a Net Dept (of a ISP with heavy use of IP routing and BGP) to check pre-depolyement of a new site... Simulation of the impact of a maintenance... etc.

    I fail to see a usage for smaller ISP that are using BGP only for peering and OSPF internaly.

     I would like to feed the full routing table, and zerba/quagga/bgp logs, or route-views data to see a play by play of evenements... and being able to inject command to try to control a evenement to reduce the impact on the net...
    (that would have been cool)

    Have fun...

Randy Bush wrote:

Why would you need to sim this at small scale?


Date: Thu, 02 Feb 2006 13:01:27 -0500
From: Alain Hebert

I'm I alone to find this a bit spammy?

Quite possibly. I for one may well be able to cross off some things

Less work? Open-source tools? (LGPL, but _c'est la vie_, I suppose.)
Far more operational than the NANOG "McDonalds" thread -- or any number
of others?

I'll take it. If there are also OSPF and IS-IS analogs of C-BGP, I'll
be _really_ happy.

Let's not slap down those who are donating useful tools to the network
engineering community.



    Just trying to categorize where one can use this...

    (Most of) The new generation of BGP experts are being formed in smaller ISP... And this is where most of the mistakes happen...

    Also thanks to Mario for pointing -> , its look promising for small scale ISP's.0 (Dont have time today to compile it... Will see in the future)

    For the other still commenting on my initial comment -> I didnt' said full on spam...

Christian Kuhtz wrote:

Alain Hebert wrote:

   (Most of) The new generation of BGP experts are being formed in smaller ISP... And this is where most of the mistakes happen...

---> Certainly not all.

A secondary point is that C-BGP could likely prove useful not only for fault detection, but also for offline traffic engineering purposes. See, for example, related algorithms and motivation in: