A long time ago, someone at Harvard cooked up two programs
running on semi-fast (at the time) PCs. The programs
were named 'hammer' and 'anvil'. Hammer was the source, anvil
was the sink. This individual then put various routers between
the two, and measured the packet rates he could obtain. In
the end he published pretty graphs, etc.
I don't remember who he was, or where to get the graphs. Perhaps
someone with more time than I can find it via archie/...
Sounds like Scott Bradner's (sob @
edu) router benchmark stuff.
Last time I asked, he'd done a few tests of filter performance,
but not many. If I'm reading the charts correctly, both Cisco and
Wellfleet routers were able to run at about full Ethernet speed
with 10 filters in place. I don't have any details of the filters
used, and in particular I don't know if he was filtering on port numbers,
or just on addresses.