[c-nsp] PPS

[Lincoln Dale (ltd)]

> > I was looking to find the total PPS the switch is doing, the switch
> > performacne sheet on cisco i think said it could handle 35,000,000
or
> > 40,000,000.  Is it talking about each interface being able to handle
> this
> > amount?  I assumed it was the total amount of everything combine?
> 
> 1GE can handle 1.488Mpps on single direction when saturated
> with small packets.
> 
> [ytti at ytti.fi ~]% echo "1000*1000*1000/(1+7+6+6+2+46+4+12)/8"|bc -l
> 1488095.23809523809523809523
> 
> So it's safe to assume, it's lookup speed for whole platform.

actually, it really does "depend" on the platform. many platforms have
'distributed' forwarding - e.g. take a 6500 or 7600.
you *can* have centralized forwarding (no DFCs installed) or you can
have 'partially centralized' (DFCs on some linecards, none on others) or
'fully distributed' (DFCs on all linecards).

although I tend to agree that 35M or 40M PPS generally sound like
"switch" platforms so are most likely centralized / whole-platform
numbers.

your math for PPS at minimum-frame-size is correct. about the only thing
I'd like to add to your math is that 'wire rate' on 1GbE can actually be
anywhere between 1487946 and 1488244 packets per second (+/- 149 PPS),
since oscillators on 1GbE are allowed to be +/- 100PPM.

if you were (say) generating traffic at 1488095 PPS, the chances are all
you'll end up showing is ever-increasing latency since you end up
showing queuing delay due to oscillator differences.  best to always
test at 99.99% of wire-rate.


cheers,

lincoln.