perfSONAR User Q&A and Other Discussion

[Tim Rayner <>]

Hi Brian & others
   >On 05/12/2014 20:39, Brian Candler wrote:
   >> But: using bwctl I am still getting lower figures than iperf3 by
   >> itself. What flags or options is bwctl running iperf3 with?

   >OK, I think I found it.
   >
   >Short version: bwctl unconditionally sets the zerocopy (-Z) option on
   >iperf3, and this is reducing the performance.

I have also noticed that the -Z flag seems to be invoked automatically, but 
not reported as such in the output results.

   >Should it just be turned off, or made a flag to bwctl?

Please don't turn it off!

With our hardware, it was the difference between achieving line rate 10Gbps 
and not.  Initially I couldn't explain why the bwctl results between hosts 
were ~>9.5 Gbps, but my manual iperf3 results were only 8.2 Gbps.  When I 
discovered the -Z option for iperf3 client it was clear that this made the 
difference..

On our hardware (supermico blade servers with "Intel(R) Xeon(R) CPU E5-2620 
v2 @ 2.10GHz" ) with integrated  ixgbe 10GigE SFP+ interface, the -Z option 
to iperf3 performs noticeably better.

I'd be happy to see the iperf3 -Z option to be made a bwctl option, or at 
least reported with the bwctl test results - it was quite confusing get 
poorer performance when I copied and pasted the reported iperf3 invocation 
from bwctl results.

Regards

Tim.