On Wed, 2004-12-22 at 08:32, Thomas Graf wrote:
> * jamal <1103721013.1090.30.camel@xxxxxxxxxxxxxxxx> 2004-12-22 08:10
> > Dont have time to look at it right now - If you can also have it
> > generate traffic in that framework and validate results that would
> > be similar to what i have.
>
> Sure, no problem. Here's an example:
>
> ts_tc "attr-police" "police creation" $FILTER police rate 2kbit buffer 10k
> drop
>
> ts_log "cls-police: Sending 10 ICMP echo requests"
> hping2 -c 10 -1 --fast ${DEST} &> /dev/null
>
> ts_tc "cls-police" "police dumping" -s -s -d filter list dev $DEV parent 10:0
>
Good start. Now if there was two hooks which could be empty. First one
to say what to do with traffic - you have it going to /dev/null
Note that i had tcpdump capture it because i wanted to analyze it for
pedit sake at the second spot: after you finish testing.
In order to really tell whether test passed or not, you need to see
those counters and analyze them too ;-> (I dont - but it is valuable).
> It could be done a lot better but I'm not aiming at replacing tcsim
> but rather to complement it so we have tcsim to validate the
> correctness of queueing and classification algorithms and my naive brutal
> testing framework for regression and stress testing under real conditions.
I think tcsim for testing the infrastructure sounds reasonable. Whats
Mr. Almesberger's long term plans for it? CCed him.
cheers,
jamal
|