[j-nsp] BFD advice

[Lewis, Charles]

We run it pretty extensively.  We've found that the timer itself really
shouldn't go much below 75 ms and that a multiplier of four has proven
safest (applied over various flavors of Ethernet within 100km or so).
When we've tried to run much tighter we've hit situations where ppmd
apparently isn't getting enough CPU time and periodic false link
failures occur.  In other words, your lab conclusions seem to be
consistent with our experiences.

There have been rumblings about putting bfd into the PFE - this would
probably yield much better detection times and improve stability.  I
don't know whether/if this is on any roadmaps, though.

I don't know if you plan to run them but there's a pretty significant
caveat with BFD on J-series boxes.  Timer values that are fine on our M
boxes have proven unstable on the J-series.  Roughly speaking we've
found stability around 800 ms or so (~200ms x 4).  We've been led to
understand that this is a function of how the kernel is prioritizing
locally terminating traffic that's specific to the platform.  Hopefully
this will be addressed soon.

CHL

-----Original Message-----
From: juniper-nsp-bounces at puck.nether.net
[mailto:juniper-nsp-bounces at puck.nether.net] On Behalf Of George Yalamov
Sent: Wednesday, August 03, 2005 11:45 AM
To: juniper-nsp at puck.nether.net
Subject: [j-nsp] BFD advice



Hi all,

	Does any one have real life experance with BFD, tunning timers, 
measurement of traffic disruption between 2 end points, convergance time

in IGP (ISIS for example).

I've tested this feature in lab environment with some M boxes, and the 
minimum time of traffic interruption was 200 - 400ms measured with
iperf.

Any ideas?


regards
George
_______________________________________________
juniper-nsp mailing list juniper-nsp at puck.nether.net
http://puck.nether.net/mailman/listinfo/juniper-nsp