rfd

Tue Dec 18 22:01:47 UTC 2018

I think you will find that very hard to evaluate since the value of RFD will be different in different network regions.  For example, it is probably good practice to run RFD toward a customer on an unstable access link.  It might not be a good idea to run it on a major backbone link that could possibly flap a large number of times in a very short period due to something like a maintenance activity.  Also, in areas that are largely on a fiber infrastructure will see RFD in a much different light than a largely wireless infrastructure that might be subject to momentary interference or interruptions.  I think it is most safe to say that RFD needs to be evaluated and tuned for what you want it to do.  Penalties are never a pleasant thing but they prevent lawlessness.  That is exactly what RFD does.  You are the cop that decides how to enforce the laws.

In fact in my experience people could also get much better network performance overall by properly tuning BGP timers but very few actually do it.  I bet you could improve the Internet stability way more by doing that.

Steven Naslund
Chicago IL

>What would really be of interest to me would be for those that run RFD to measure its impact to their network (positive or otherwise) so we have something scientific to base on.
>
>The theory (and practice of old) tells us that RFD is either very good, or very bad. There are probably more folk that have turned it off than run it, or vice versa. Ultimately, if we can get the state >of RFD's performance in 2018 on an axis, our words will likely carry more weight.
>Mark.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.nanog.org/pipermail/nanog/attachments/20181218/189909b0/attachment.html>