Ran  iperf on straight linux vxlan  interfaces between the same 2 hosts.   
In that case the I see a throughput  of 7.04 gbps

My guess at this point is  the  3.03 Gbps reduction is caused by linux 
bridge in the overlay space.  Would appreciate any insight. 



On Thursday, September 15, 2016 at 2:47:31 PM UTC-7, sgas wrote:
>
> Thanks, rerun the  iperf test at higher MTU . 
>
> MTU: 8950 :  4.01 Gbits/sec
>   
>
> On Thursday, September 15, 2016 at 9:58:20 AM UTC-7, sgas wrote:
>>
>> The following are some  numbers I observed with iperf across  containers 
>> on 2 hosts . 
>>
>> MTU  1450  :  1.11 Gbits/sec
>>
>> A bigger MTU   obviously will increase the throughput  .    Here are some 
>>  questions that anyone probably would ask at this point. All these are in 
>> the context of overlay networks. 
>>
>> 1.   What is the timeline to support  bigger mtu ( provided underlying 
>> infrastructure supports ) .
>> 2.   Is there a plan to  use vxlan offload if supported by the NIC ? 
>> 3.   Any thoughts on mix mtu  environment  ( e.g  say one  container  at 
>> mtu 1450 and the other at 8950 ) ? 
>> 4.   Any thoughts on  interacting with other  devices  that support 
>> vxlan/vtep (e.g  switches , routers ) ?
>>
>>
>>
>>
>>

-- 
You received this message because you are subscribed to the Google Groups 
"docker-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to docker-dev+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to