[netperf-talk] global question concerning Netperf test and SMP support

[Rick Jones]

On 04/27/2012 04:05 AM, Simon Duboue wrote:
> Hello,
>
>  >These were consecutive (one at a time) right? Because if they were
>  >concurrent (all at once) you really must set max,min iterations for the
>  >confidence intervals to the same value, lest one or more of the
>  >instances terminate before the others.
>
> They are normally concurrent. I think I just forget to put the -P 0
> options in what I echo. I execute with -P 0 and I echo without. Max,min
> iteration are set to the same value (-i 10,2 -I 99,5) so I don't really
> understand your remark... Could you explain more?

I did not see '&' at the ends of your command lines.  If I were to enter 
those command lines as they appeared in the email, they would have run 
one at a time.

>  >The output as you have presented it makes it look as though the tests
>  >ran one-at-a-time.
>  >
>  >Another sanity check would be to compare the service demands between a
>  >single (all by itself) test and then with your five-at-once tests.
>  >Since each netperf/netserver measures overall CPU utilization, but only
>  >what it itself transferred, when you run actually concurrent tests, the
>  >service demands reported by the ostensibly-all-at-once tests will be
>  >(much) higher (and wrong :). If you see the same service demand for
>  >each of the ostensibly-all-at-once as you do for a single test, you know
>  >the ostensibly-all-at-once were not really all-at once.
>
> Ok, this is a good complementary check. Here is the result of the first
> test I do concerning service demand:
> single test:
> netperf -l 60 -H 10.0.17.200 -i 10,2 -I 99,5 -t TCP_STREAM -cC
> TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 10.0.17.200
> (10.0.17.200) port 0 AF_INET : +/-2.5% @ 99% conf. : demo
> Recv Send Send Utilization Service Demand
> Socket Socket Message Elapsed Send Recv Send Recv
> Size Size Size Time Throughput local remote local remote
> bytes bytes bytes secs. 10^6bits/s % S % S us/KB us/KB
>
> 87380 65536 65536 60.01 2075.12 *5.19 9.32 0.409 3.027 *
>
>
> two-at-once test:
> netperf -l 60 -H 10.0.17.200 -i 10,2 -I 99,5 -t TCP_STREAM -cC &
> netperf -l 60 -H 10.0.17.200 -i 10,2 -I 99,5 -t TCP_STREAM -cC
>
> TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 10.0.17.200
> (10.0.17.200) port 0 AF_INET : +/-2.5% @ 99% conf. : demo
> TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 10.0.17.200
> (10.0.17.200) port 0 AF_INET : +/-2.5% @ 99% conf. : demo
>
> Recv Send Send Utilization Service Demand
> Socket Socket Message Elapsed Send Recv Send Recv
> Size Size Size Time Throughput local remote local remote
> bytes bytes bytes secs. 10^6bits/s % S % S us/KB us/KB
>
> 87380 65536 65536 60.06 1760.01 *9.21 13.11 0.858 4.801 *
>
> Recv Send Send Utilization Service Demand
> Socket Socket Message Elapsed Send Recv Send Recv
> Size Size Size Time Throughput local remote local remote
> bytes bytes bytes secs. 10^6bits/s % S % S us/KB us/KB
>
> 87380 65536 65536 60.06 1750.31 *9.21 13.14 0.861 4.730*
>
> This looks good isn't it?

That definitely looks better.  However, you need to change the "-i 10,2" 
to "-i 10"  You must not let netperf finish "early" if it hits the 
confidence interval, so the min and max iterations must be the same.

rick