On Sun, Mar 30, 2008 at 10:03 PM, Tony Li <[EMAIL PROTECTED]> wrote: > There hasn't been much discussion on the mapping churn rate. There have > been a few good questions, but not much back and forth. Do folks have an > opinion or should I just propose some text?
Hi Tony, I think there may be four relevant numbers: Maximum per-machine RIB-style update rate, that is map updates which are conveyed by the machine but not used in the machine's forwarding plane Maximum per-machine FIB update rate for TCAM-based FIBs Maximum per-machine FIB update rate for hardware-accelerated trie-based FIBs Maximum per-machine FIB update rate for non-accelerated trie-based FIBs (read: generic x86 server) I have trouble imagining a system that doesn't have to contend with some or all of these limits. No matter how cleverly we structure it, our system will still be composed of individual machines. Call me crazy, but I think we pretty much understand how RIBs and FIBs work within the scope of an individual machine and we know what we can expect with hardware available today. We can punch BGP routes into Quagga unbound from server route table and see where it chokes to get a decent estimate of the maximum per-machine RIB update rate. Similar strategies should work for the other three. Whether the knowledge source is a map or a route doesn't make much of a difference at the individual-machine level. Or am I wrong? I don't think there is a useful way to pre-assess system-wide maximums aside from asking proposal authors to specify how tightly bound the system-wide maximum is to the heaviest-loaded machine maximum and then justify that analysis. 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
