[openstack-dev] [Fuel] 10Gbe performance issue.

Aaron Rosen aaronorosen at gmail.com
Tue Jan 20 18:04:25 UTC 2015


If you can get 9Gbps with multiple connections I'm guessing it's because of
latency and the buffer size of your sockets. If you change the sending and
receiving window size you should be able to fully saturate the link with
one connection (though there are several reasons for not doing that).

On Tue, Jan 20, 2015 at 8:14 AM, Piotr Korthals <piotr.korthals at intel.com>
wrote:

>  Hi,
>
> I am facing issue with performance of 10 Gigabit Ethernet.
>
> Environment 2 hosts each:
> - 2 * Intel(R) Xeon(R) CPU E5-2690 v2
> - Intel 82599ES 10Gb Ethernet
> - 128GB RAM
>
> System:
> - Centos 6.5 delivered by fuel 6.0
>
> iperf during test of network performance over single stream report
> 2,5-3Gbps (when using 4+ connection, cumulative transfer can hit over 9Gbps)
>
> For comparison, same test was run on official Centos 6.5 with result at
> minimum 9,3 Gbps
>
> In both cases default MTU 1500 was set.
>
> From my analyses it looks like something is limiting i/o performance per
> stream.
>
> Are you aware of this issue, and can you comment on it?
>
> This is critical for our deployment.
>
>
> __________________________________________________________________________
> OpenStack Development Mailing List (not for usage questions)
> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20150120/8ed6599b/attachment.html>


More information about the OpenStack-dev mailing list