The task at hand is to figure out whether "mls ip cef load-sharing full" would make a good use of L4 ports. "Good use" would produce, from randomly chosen L4 ports, a reasonably uniform distribution of output interfaces.
Platform is: IOS 12.2(18)SXE4 under 7609, experimenting with a pair of parallel p2p output links.
Cisco's command reference about "mls ip cef load-sharing full" states:
full (Optional) Sets the CEF load balancing to include source and destination Layer 4 ports and source and destination IP addresses (Layer 3).
However, experiments are leading me to believe that changes in L4 ports scarcely affect the selection of the output interface. That is, for a fixed pair of source-destination addresses, most changes of (either source or destination) L4 ports keep yielding the same output interface.
Can anyone confirm this, or shed some light on why changes in L4 ports seem to produce such uneven distribution of output links?
Thanks, Everton