On Tue, Mar 25, 2008 at 8:31 PM, Brian E Carpenter <[EMAIL PROTECTED]> wrote: > But if we fix that by having the ETR send a map update (actually > a map withdrawal) and we need that to propagate in less than the > TCP timeout, haven't we just turned the mapping system into > a dynamic routing protocol? However, it has to handle > several orders of magnitude more entries than BGP.
Hi Brian, Well, yes. Isn't that the whole idea? We have this large and growing FIB and it's hideously expensive any time we have to implement it on traffic flows which are larger than what a generic x86 can handle. So we move the bulk of that "routing" out to ITRs at the edge where the traffic flows are small and the only thing we keep in the high-rate FIBs is the knowledge necessary to get from the ITRs to the ETRs. I mean, we get one, possibly two orders of magnitude on the RIB just by moving it to multicore servers with lots of ram, devices in the $8k range. RIB growth may be superlinear but it's slower than Moore's Law. The bottleneck is the FIB, specifically on devices with a larger-than-COTS traffic flow. Right? Regards, Bill Herrin -- William D. Herrin ................ [EMAIL PROTECTED] [EMAIL PROTECTED] 3005 Crane Dr. ...................... Web: <http://bill.herrin.us/> Falls Church, VA 22042-3004 -- to unsubscribe send a message to [EMAIL PROTECTED] with the word 'unsubscribe' in a single line as the message text body. archive: <http://psg.com/lists/rrg/> & ftp://psg.com/pub/lists/rrg
