Yu Ning <yuning@ns.chinanet.cn.net> writes:
> FIB (Forward Information Base) and Adjacency Database (AD)
> is component of CEF. In CCO document, it's said that FIB
> is derived from RIB, can any one tell me why the FIB+AD
> architecture is faster than traditional RIB+ARP ?
Yes - there are two reasons:
1) CEF pre-builds the RIB such that all available prefixes are
available without having to process-switch the first packet
for a particular prefix to populate the route-cache as with
legacy fast-path mechanisms.
2) The CEF table does not need to be invalidated when the next-
hop link-layer address changes, since the CEF table contains
pointers to the Adjacency database (which is pre-built with
next-hop->Media-address data and updated separately from the
CEF table) rather than looking up to the next-hop->Media-
addresses.
> A 256-way M-Trie structure is used with a patented
> algorithm to find matches with at most 4 lookups.
As I understand it, the 256-way tree structure algorithm
is essentially similar to the way the 'optimum' switching
mode works. The fundamental difference being that the
lookup is to a pointer into the adjacency table rather
than the next-hop information itself.
M.
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:16 EDT