Progressive BGP route dampening
This is a short note to the ripe community and our peers to let you know that we are now implementing a form a progressive route dampening based on BGP prefix length. Flapping routes with longer prefixes will be penalised/suppressed for longer than shorter ones. Hopefully more implementations of this ilk will prompt more Networks to CIDRise. Maximum ripe block allocations will receive the default cisco suppression. Any questions, just drop us a line. Regards --Tony
Tony, you might want to be a bit more specific about the values you are using, i.e. how long certain prefixes will be ignored. This will make troubleshooting easier and reduce calls into your NOC. This brings up an interesting question too: Can we still troubleshoot the connectivity problems arising if transit ISPs use all different values for this or even the proposed exponential version? Daniel
Tony Barber <tonyb@pipex.com> writes:
This is a short note to the ripe community and our peers to let you know that we are now implementing a form a progressive route dampening based on BGP prefix length. Flapping routes with longer prefixes will be penalised/suppressed for longer than shorter ones. Hopefully more implementations of this ilk will prompt more Networks to CIDRise.
Maximum ripe block allocations will receive the default cisco suppression.
Any questions, just drop us a line.
Regards
--Tony
I'd like to second Daniel's suggestion
you might want to be a bit more specific about the values you are using, i.e. how long certain prefixes will be ignored. This will make troubleshooting easier and reduce calls into your NOC.
I think it would be helpful to publish those parameters both to operators' lists in e-mail and register them in remarks in the AS objects thru the RRs. I don't know if there is any need to have someone more knowledgable to propose some specific syntax and parameters. Ruediger Volk ### this .signature is currently under construction ### Deutsche Telekom AG -- Internet Services NIC E-Mail: rv@NIC.DTAG.DE
Daniel Karrenberg wrote:
Tony,
you might want to be a bit more specific about the values you are using, i.e. how long certain prefixes will be ignored. This will make troubleshooting easier and reduce calls into your NOC.
This brings up an interesting question too: Can we still troubleshoot the connectivity problems arising if transit ISPs use all different values for this or even the proposed exponential version?
Daniel
Well currently /24 and greater for 'around' 3 hours /22 /23 for about an hour everything else defaults to about 25 minutes. But it depends on the flap frequency. Continual flapping will drive the times up untill they hit an upper limit of 255 for /24+ and 90 minutes for /22 and /23 Is there a possibility that we could have a discussion on this at the next RIPE and possibly agree pan-european values to use ? --Tony
Der Tony, you wrote:
Message-Id: <199607160940.KAA28911@pipe.pipex.net> Subject: Re: Progressive BGP route dampening To: Daniel.Karrenberg@ripe.net (Daniel Karrenberg) Date: Tue, 16 Jul 1996 10:40:27 +0100 (BST) Cc: eof-list@ripe.net, routing-wg@ripe.net From: Tony Barber <tonyb@pipex.com>
Daniel Karrenberg wrote:
you might want to be a bit more specific about the values you are using, i.e. how long certain prefixes will be ignored. This will make troubleshooting easier and reduce calls into your NOC.
This brings up an interesting question too: Can we still troubleshoot the connectivity problems arising if transit ISPs use all different values for this or even the proposed exponential version?
Well currently /24 and greater for 'around' 3 hours /22 /23 for about an hour everything else defaults to about 25 minutes.
But it depends on the flap frequency. Continual flapping will drive the times up untill they hit an upper limit of 255 for /24+ and 90 minutes for /22 and /23 Thank you for this specific information. How did you arrive at these values? I estimate you had more than one thought on this...
Is there a possibility that we could have a discussion on this at the next RIPE and possibly agree pan-european values to use ? I am definitely interested in a discussion. It would be nice to have a report from you on your experience with progressive BGP route dampening. May I ask you to prepare something for the next RIPE meeting? Regards Joachim
DFN Network Operation Center, Dr. Joachim Schmitz, (finger help@noc.dfn.de) >>>>>> mailto: noc@noc.dfn.de <<<<<< Rechenzentrum Universitaet Stuttgart, Allmandring 30, D-70550 Stuttgart Phone: 0711-685-5810, 0711-685-5576, FAX +711 6788363 (business hours) EMERGENCY phone +711 1319 139 with answering machine and pager _____________________________________________________________________________
Tony Barber <tonyb@pipex.com> writes: Is there a possibility that we could have a discussion on this at the next RIPE and possibly agree pan-european values to use ?
Certainly. I am not sure whether it belongs into EOF or Routing WG. But if you ask both chairpeople, they will work it out together. Daniel
participants (4)
-
Daniel Karrenberg -
Ruediger Volk -
Schmitz＠RUS.Uni-Stuttgart.DE -
Tony Barber