[c-nsp] Switching paths

[Reuben Farrelly]

I'm interested in an aspect of the switching technologies used in routers, 
specifically which features are and aren't well supported in each switching 
path.  In the case I am thinking of right now I'm thinking of an 1841 and 2851 
which I recently configured for an end customer.

http://www.cisco.com/warp/public/765/tools/quickreference/ gives performance 
figures for CEF and process switched tests and states that "If a feature is 
supported in the CEF path..."   But it doesn't say how we would find out if a 
feature is actually supported in CEF or not ;-)

I'm interested in these features specifically and what effect they might have on 
the switching performance of a router:

* Crypto maps
* CBAC content inspection
* IPS/IDS
* ACL's
* ACL's with 'log' as one of the keywords, does this have an impact if the match 
is achieved before the log statement?
* WCCP
* NAT - which I think may have been fully supported in the CEF path for a fair while
* GRE tunnels with crypto maps applied

Thinking specifically of 12.4/12.4T here as those are the only ones that run on 
the new ISR's.

Is there a relatively easy way to find out from a device, what is forcing 
traffic to be punted to a non-CEF path if CEF is configured?  Or a web page on 
CCO which says which features are fully CEF supported?  (I couldn't find one)

Yes I've read 
http://www.cisco.com/en/US/customer/products/sw/iosswrel/ps1828/products_tech_note09186a00801e1e46.shtml 
but that doesn't really explain how to find out which feature is forcing traffic 
to be punted to a non-CEF path, nor what sort of percentage of CEF/non-CEF 
traffic is 'acceptable' if all the features are supported in the CEF path (I've 
seen ~30% traffic process switched and I'm thinking that's a bit high).

Thanks,
Reuben