[ovs-discuss] performance

edward wilson edward.wilson2601 at gmail.com
Tue Sep 11 17:44:49 UTC 2012


Hi

Means for actual performance testing, I have to test it on real
environment. But if I want to measure the  increase in performance from
tool then how can I measure it.

Is ovs-benchmark is ok to measure the increase in performance between the
versions?
if not then please suggest me any other tool

Regards
Ed

On Tue, Sep 11, 2012 at 10:05 PM, Jesse Gross <jesse at nicira.com> wrote:

> On Mon, Sep 10, 2012 at 11:52 AM, edward wilson
> <edward.wilson2601 at gmail.com> wrote:
> > Hi,
> >
> > I have used ovs-benchmark and found drastic change everytime I run the
> > command with same parameter and environment.
> >
> > For comparing between older and latest version of OVS,
> > With older version over 10 iterations average rate with timeout 5 comes
> out
> > to be 2680.92 whereas on latest OVS version it is 3826.34 .It is just 1.4
> > times
> > With older version over 10 iterations average rate with timeout 1 comes
> out
> > to be 3907.97 whereas on latest OVS version it is 5245.10 .It is just 1.3
> > times
> >
> > Please confirm how this performance is triple as this is mentioned that
> flow
> > setup performance is tripled in latest version ?
> >
> > Is ovs-benchmak ok for this measurement?
> > How can we ensure tripled performance in latest vesion of OVS?
>
> ovs-benchmark is a synthetic benchmark.  If you are concerned about
> the impact of performance changes in newer releases you should test
> with your workloads.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://openvswitch.org/pipermail/ovs-discuss/attachments/20120911/3498bbdf/attachment.html>


More information about the discuss mailing list