Date: Mon, 9 Dec 2002 14:50:18 +0100 From: "Koepp, Karsten" <Karsten.Koepp@lambdanet.net>
can anyone of you point me to statistics about the frequency of BGP updates measured at some point in the internet? A customer complained about av. 10 updates per second which we consider normal, but we'd like to see values for comparison.
Karsten, thats really an interesting question: ACOnet receives a full routing-table from SPRINT/AS1239 and from GEANT/AS20965. Comparing the current BGP session counters: neighbor SPRINT, uptime ~20weeks, Received 7926940 messages neighbor GEANT, uptime ~ 7weeks, Received 29597194 messages which gives a long-term average of 419 BGP update messages per minute from GEANT but only 39 BGP update messages per minute from SPRINT afaik GEANT does no BGP flap damping, SPRINT however does, according to http://www.sprint.net/policy/bgpdamp.html I don't know if damping is the only reason for the factor >10 between them, there might be other reasons as well, e.g. SPRINT backbone is Cisco based, GEANT is Juniper based and they might be using completely different IGPs and iBGP concepts. I am also suspecting some improperly behaving BGP implementations beeing around, e.g. at the VIX we are having peering neighbors who are sending superproportional more BGP updates than others with the majority of the messages beeing withdrawals for prefixes which they do not own nor announce. Btw, Lambdanet is sending us a long-term average of 2 BGP update messages per minute at the VIX. Kind regards CP --- ---------------------------------------------------------------------- --- --- Christian Panigl : Vienna University Computer Center - ACOnet --- --- VUCC - ACOnet - VIX : -------------------------------------------- --- --- Universitaetsstrasse 7 : Mail: Panigl@CC.UniVie.ac.at (CP8-RIPE) --- --- A-1010 Vienna / Austria : Tel: +43 1 4277-14032 (Fax: -9140) --- --- ---------------------------------------------------------------------- ---