On Apr 16 20:58, Donnelly, John wrote:
> > how did you invoke the iperf server instance?
> 
> I run ssh sessions between end points :
> 
> servers:
> 
>   ssh -f  -l root 10.10.50.2     iPerf -S -d & 
>   ssh -f  -l root 10.10.60.2     iPerf -S -d &
> 
> clients :
> 
>   ssh -f -l root 10.10.60.2 iperf -f M -t 61 -c 100.100.50.11 &
>   ssh -f -l root 10.10.60.2 iperf -f M -t 61 -c 100.100.50.12 &
>   ssh -f -l root 10.10.60.2 iperf -f M -t 61 -c 100.100.50.13 &
>   ssh -f -l root 10.10.60.2 iperf -f M -t 61 -c 100.100.50.14 &
>   ssh -f -l root 10.10.60.2 iperf -f M -t 61 -c 100.100.50.15 & 
>   ssh -f -l root 10.10.60.2 iperf -f M -t 61 -c 100.100.50.16 & 
>   ssh -f -l root 10.10.60.2 iperf -f M -t 61 -c 100.100.50.17 &
>   ssh -f -l root 10.10.60.2 iperf -f M -t 61 -c  100.100.50.18 &
> 
>   ssh -f -l root 10.10.50.2 iperf -f M -t 61 -c 100.100.60.11 &
>   ssh -f -l root 10.10.50.2 iperf -f M -t 61 -c 100.100.60.12 &
>   ssh -f -l root 10.10.50.2 iperf -f M -t 61 -c 100.100.60.13 &
>   ssh -f -l root 10.10.50.2 iperf -f M -t 61 -c 100.100.60.14 &
>   ssh -f -l root 10.10.50.2 iperf -f M -t 61 -c 100.100.60.15 & 
>   ssh -f -l root 10.10.50.2 iperf -f M -t 61 -c 100.100.60.16 & 
>   ssh -f -l root 10.10.50.2 iperf -f M -t 61 -c 100.100.60.17 &
>   ssh -f -l root 10.10.50.2 iperf -f M -t 61 -c  100.100.60.18 &
> 
>   The Linux 5.4 machines have multiple ip addesses on a single adapter 
>     server 1  <10.10.50.2>             server 2 < 10.10.60.2 >
>     ifcfg_eth1:1    100.100.50.11         100.100.60.11
>               :2    100.100.50.12         100.100.60.12
>               :3
>              <n ports>                     <n ports>

Hi John,

I'm trying to understand your goal here - is there a specific use case, are you
attempting to validate iperf functionality, or something else entirely?  Not
poking, I just can't see the benefit of the test scenario you've described.
Since all of the interfaces are essentially the same, why not just invoke the
client with a specific number of threads, e.g. -P 8?

> >> or  non-C++ Linux version available ?
> 
> > Frankly, I am confused by this question...
> 
>  A non C++ (non-threaded) version. 

Ah, mmkay.  You have a couple of options:

1) recompile w/out threading support (./configure --disable-threads)

2) invoke iperf with a single thread (-P 1); default is to run threaded if
support is compiled in.

So, in your test scenario (if truly needed), you could run mulitple server
instances, each with a single thread and each on their own port.  I suspect
what you're seeing is actually a clash between the multiple client instances
you've spawned (vs. invoking with a specific number of threads), but that's
just a guess.

I will note that I can get the server to report different finish times if I
invoke multiple successive clients with some time in between each:

[ 11]  0.0-61.1 sec    280 MBytes  4.57 MBytes/sec
[  4]  0.0-66.9 sec    221 MBytes  3.30 MBytes/sec
[  5]  0.0-73.2 sec    201 MBytes  2.74 MBytes/sec
[  6]  0.0-101.3 sec    435 MBytes  4.29 MBytes/sec
[SUM]  0.0-101.3 sec  1136 MBytes  11.2 MBytes/sec

...but again, invoking in that way seems to be unnecessary.

For giggles, try running your client against a _single_ server address with
multiple client threads, and see what you get.

Cheers,

-tt


-- 
Tom Throckmorton
MCNC - Advanced Services Development
3021 Cornwallis Road
Research Triangle Park, NC 27709
919.248.1448

"Connecting North Carolina's future today"

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
_______________________________________________
Iperf-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/iperf-users

Reply via email to