[nsp] 7500: interaction of dCEF and "ip accounting"

[hakan lindholm]

Hello all.

Is it really true that dCEF gets disabled by ip accounting?

I have a 7507-MX with this config:

interface FastEthernet1/1/0
 ip accounting mac-address input
 ip accounting mac-address output

IOS version is 12.0(25)S1,
RSP is 4,
VIP's are 4-80 and GEIP,
PA's are FE-TX and POS-OC3-LR.


I get all the expected in/out counters for "peers" with SNMP.

Maybe "ip accounting out" is different from "ip accounting mac-address
out"

/Hakan



On Sat, 17 Jan 2004, Denis V. Schapov wrote:

> Hi.
> 
> "Note: IP accounting disables DCEF. Do not enable IP accounting on any platform where you need to do DCEF switching."
> http://www.cisco.com/warp/public/63/ts_codred_worm.shtml
> 
> And some others urls on CCO that casually note this fact - you can get it through search
> 
> Restriction was documented through CSCeb05478
> http://www.cisco.com/univercd/cc/td/doc/product/software/ios122/122cgcr/fipras_r/1rfip1.htm#1018848
> 
> "IP accounting disables autonomous switching, SSE switching, and distributed switching (dCEF) on the interface. IP accounting will cause packets to be switched on the Route Switch Processor (RSP) instead of the Versatile Interface Processor (VIP), which can cause performance degradation."
> 
> Netflow feature accelerate doesn't help here. It was also tried.
> Also you can consider table for switching path (switching on incoming/outgoing interface and resulting switching path)
> http://www.cisco.com/warp/public/105/cef_whichpath.html#verify
> 
> 
> > Hi,
> > 
> > could someone point me at some documentation describing the interactions
> > of dCEF (on a 7500) and "ip accounting out"?
> > 
> > I know that dCEF won't work with "ip accounting" (at least that's what
> > I've heard, but maybe it *does*, if "netflow feature-accelerate" is
> > switched on, as netflow *does* work distributed?).  
> > 
> > Anyway, assuming it doesn't work together.
> > 
> > Now we have a 7500 that has "ip accounting out" on *some* interfaces, but
> > not *all*.  So we can have packets flowing
> > 
> >  VIP with "ip accounting out" -> VIP with "ip accounting out"
> >  VIP witout "ip accounting o" -> VIP with ...
> >  VIP with ...                 -> VIP without ...
> >  VIP without ...              -> VIP without ...
> > 
> > Now which of those paths can be dCEF accelerated, and which paths need
> > to go to the RSP for "central CEF"?
> > 
> > Naively I'd assume that all paths that have a destination interface
> > without "ip accounting out" could be dCEF switched on the VIPs, but the
> > interactions might be more complex...
> > 
> > Current IOS is 12.0(21)S1 on a 7507, RSP4s, VIP2-40 and VIP2-50s (plus an 
> > old FSIP-8, which cannot do distributed <anything>, of course).  
> > 
> > Eventually the box will move to 12.2(x)S to get IPv6.
> > 
> > thanks!
> > 
> > gert
> > -- 
> > USENET is *not* the non-clickable part of WWW!
> >                                                            //www.muc.de/~gert/
> > Gert Doering - Munich, Germany                             gert at greenie.muc.de
> > fax: +49-89-35655025                        gert at net.informatik.tu-muenchen.de
> > _______________________________________________
> > cisco-nsp mailing list  cisco-nsp at puck.nether.net
> > https://puck.nether.net/mailman/listinfo/cisco-nsp
> > archive at http://puck.nether.net/pipermail/cisco-nsp/
> > 
> 
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/
>