*Hi Jesse,* Yes, I see your point that there is no benefit of controlling incoming traffic, because anyways all traffic will arrive atleast till OVS level from physical NIC, so that channel will be the bottleneck anyway.
I configured OVS using Xen as Hypervisor, and, I am able to create VMs which automatically gets connected to the bridge that OVS created. But I am not seeing any QoS control in the VM. Here is the bridges and interfaces, cadlab:~ # brctl show bridge name bridge id STP enabled interfaces eth0 0000.7071bc62737a no peth0 vif6.0 vif8.0 I configured QoS on one of the VMs like this: cadlab:~ # ovs-vsctl set Interface vif8.0 ingress_policing_rate=1000 cadlab:~ # ovs-vsctl set Interface vif8.0 ingress_policing_burst=100 VMs are as follows : cadlab:~ # xm list Name ID Mem VCPUs State Time(s) Domain-0 0 6012 8 r----- 2271.1 opensuse11 8 1024 1 -b---- 15.3 opensuse11-clone 6 1024 1 -b---- 21.4 Then I ran netperf test, I don't see max of 1 Mbps !!! vm1@linux-g9jl:~> netperf -H 10.112.10.35 TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 10.112.10.35 (10.112.10.35) port 0 AF_INET : demo Recv Send Send Socket Socket Message Elapsed Size Size Size Time Throughput bytes bytes bytes secs. 10^6bits/sec 87380 16384 16384 10.02 93.46 I see 93 Mbps !!! Why is it so? Does anybody can tell me where I am wrong? * ---------------------------- Thanks & Regards Mohit Dhingra +919611190435* On 9 February 2012 22:31, Jesse Gross <je...@nicira.com> wrote: > On Wed, Feb 8, 2012 at 11:24 PM, Mohit Dhingra <mohitdhing...@gmail.com> > wrote: > > The question that I was asking regarding monitoring of incoming traffic > into > > VMs is, we have one ingress_policing_rate in ovs_vsctl for an interface, > but > > there is no egress/outgress policing rate.. Does that mean that it > doesn't > > have control over incoming traffic onto VMs? > > Egress queuing is possible (and is the most common use case when > applied to a physical interface) - look at the QoS section of the > ovs-vsctl man page. However, it's unlikely to be particularly useful > when applied to VMs. It won't reduce link usage at all because the > traffic has already arrived and it's unlikely to reduce CPU usage very > much single QoS is expensive to perform. >
_______________________________________________ discuss mailing list discuss@openvswitch.org http://openvswitch.org/mailman/listinfo/discuss
