this shouldn't be the case. provided that you can download the
NAT translation table entry into the PXF path, the NAT translation
is done without a perfomance hit.
some useful commands to troubleshoot what is happening would be:
show pxf info
show pxf feature nat stat
show pxf feature nat entry
show pxf acc sum
show ip nat stat
show proc cpu
it may be (if you are hitting the 30 kpps @ 320 megs) you are hitting
the limitation of the PA-GE. the PA-GE is a connectivity solution
and does not give you wire rate performance. the IO based GE interfaces
and the GE interfaces on the 7400 can give you much higher performance.
regards
.siva
On Mon, 9 Jul 2001, Mikael Abrahamsson wrote:
> On Mon, 9 Jul 2001, George Robbins wrote:
>
> > Sort of. The general rule is that it doesn't speed up anything useful
> > in the ISP realm, but can do magic in the content provider or firewall
> > domain where you care what's beyond the packet header. The main focus
> > seems to be on that and then traffic management.
> >
> > see: http://www.cisco.com/warp/public/cc/pd/ifaa/prossor/nse1/
>
> We pushed 320 megabit of traffic thru a 7206VXR with PFX. It was NATing
> the traffic at the time. PA-GEs were used. This was with very little
> traffic but 10 highspeed TCP sessions going thru the machine.
>
> With a lot of people doing online gaming (CS mostly) it seemed the
> performance was about halfed, so I guess it's the PPS that really hurts.
>
> 28-30kpps (probably someone flooding) made the routers totally bogged
> down. I guess it's around here the 720xVXR reaches its limit?
>
> --
> Mikael Abrahamsson email: swmike@swm.pp.se
>
>
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:44 EDT