Have you considered using sFlow? Its a widely implemented data plane
telemetry standard (supported by most merchant silicon) based on
random sampling of packets. The instrumentation is independent of the
control plane which has advantages. Composing OpenFlow measurement and
forwarding rules is complex and if you were to push measurements over
the same OpenFlow channel used to control the network device you would
risk flooding the channel and causing control messages to be delayed
or lost, possibly losing control of the device. Separation of
telemetry and control allows both to scale independently.

If you are prototyping with Mininet, you can use the sFlow
instrumentation built into Open vSwitch.

In much the same way that OpenFlow shifts intelligence to an external
controller, sFlow shifts analytics to an external analyzer. For your
use case, the sFlow collector would receive samples of all traffic
flowing through the network and it is up to the collector to aggregate
the sampled packet into flows by domain.

The following articles provides additional information:
http://blog.sflow.com/2011/05/openflow-and-sflow.html
http://blog.sflow.com/2013/08/restflow.html

On Fri, Mar 4, 2016 at 5:10 AM, APOORV SHUKLA
<apoorv.shukla1...@gmail.com> wrote:
> Hi Folks,
>
> I am just trying to do sampling in openflow like sending one out of N
> packets entering a network domain (it can be any flow and not just
> microflow) to any collector or controller. How can I achieve that using flow
> rules just by match action paradigm ? Can anybody suggest please ?
>
> I am trying to do it through prioritising some rules.
>
> Appreciate all help! Thanks!
>
> Warm Regards,
> -Apoorv Shukla
>
> "A great oak is only a little nut that held its ground"
>
>
>
>
>
> _______________________________________________
> openflow-discuss mailing list
> openflow-discuss@lists.stanford.edu
> https://mailman.stanford.edu/mailman/listinfo/openflow-discuss
>
_______________________________________________
openflow-discuss mailing list
openflow-discuss@lists.stanford.edu
https://mailman.stanford.edu/mailman/listinfo/openflow-discuss

Reply via email to