On Tuesday 10 March 2009 01:21:35 pm Richard A Steenbergen wrote: > Meanwhile, my RE's are constrained at 100% CPU every > single time there is any instability or reconvergence > event in the network (or sometimes they're at 100% cpu > 24/7 depending on how many unresolved rpd bugs I'm being > affected by at any given moment :P). Given those > realities, I'd gladly trade memory for better performance > any day of the week.
In one of the IS-IS tutorials I've held at past community events/workshops, we look at limiting exposure of the control plane CPU by taking advantage of some of the throttling capabilities available in vendor routers implementing the popular link state routing protocols today, i.e., mainly SPF and PRC delay. As Chris already mentioned, even in the largest of networks, full SPF runs take only tens of milliseconds to complete, and generally, events that set off iSPF or PRC do not occur that often. Juniper have the 'spf-delay' command, while Cisco have the 'spf-interval' and 'prc-interval' commands. The algorithms between both vendors differ, but the end goal is the same. These are particularly useful during times of "badness". The trick is to have values that are high enough for routers not to suffer during heavy instability, while still keeping them low so that reconvergence happens in a timely manner. Cheers, Mark.
signature.asc
Description: This is a digitally signed message part.
_______________________________________________ juniper-nsp mailing list [email protected] https://puck.nether.net/mailman/listinfo/juniper-nsp
