[ovs-discuss] monitoring traffic VM using sflow

[Peter Phaal]

The sFlow reported utilization numbers should be accurate for the physical NICs. 

There is no well defined link capacity for a vNIC and so the utilization numbers are meaningless (I believe the Open vSwitch sFlow implementation reports a nominal 100Mbits/s capacity for vNICs).  When examining vNIC interface statistics, you should focus on bits/second, packets/second metrics etc. and ignore utilization.

CPU utilization metrics for the hypervisor are a better measure of how close the vSwitch is to capacity and of virtual network utilization. 

http://blog.sflow.com/2011/10/comparing-sflow-and-netflow-in-vswitch.html

You can monitor the CPU along with the network traffic if you install a Host sFlow agent:

http://host-sflow.sourceforge.net/

Peter

On Oct 23, 2012, at 8:52 AM, Giz eudocia <gizeudocia at ymail.com> wrote:

> hello, I want to ask, why the sFlow flow statistics from interface counter value reaches a high number (frame / s) and why the presentation interface utilization exceeds 100%? this happens when I'm doing port scanning inter-VM measurements, thanks
> _______________________________________________
> discuss mailing list
> discuss at openvswitch.org
> http://openvswitch.org/mailman/listinfo/discuss

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://openvswitch.org/pipermail/ovs-discuss/attachments/20121024/f86b4369/attachment-0001.html>