advertised media rate does not equal actual link rate. UCS vnic's advertise as 40gbit's they are barely able to push 7Gbit. For example.
On 9 May 2016 at 16:50, Edward Bond <celpa.f...@gmail.com> wrote: > The one you are referencing is to the loopback device on the VM, not VM to > VM. > > recap: > VM loopback device on VM = 26 gbps > VM traffic is limited to 10gbs ( actual data is ~5 per 10G link on host ) > Host to Host is 18 gbs ( actual data is ~9 per 10G link on host ) > Host bond0 = 20gbs link speed > **Host Openvswitch link speed = 10gbs > > > ** This is the one I am most curious about > > > On Sun, May 8, 2016 at 10:56 PM, Scott Lowe <scott.l...@scottlowe.org> > wrote: > >> Please see my responses inline, prefixed by [SL]. >> >> >> > On May 8, 2016, at 4:35 PM, ed bond <celpa.f...@gmail.com> wrote: >> > >> > Scott, >> > >> > I agree. I am not expecting that. >> > >> > <snip> >> > >> > When I noticed Scenario 1, I looked at the openvswitch virtual ethernet >> device, it only has 10gbs set to the link speed. That’s when I figured I >> would send out the question. >> >> >> [SL] Sorry, I'm unclear then. Is your question how to increase the link >> speed setting for the virtual Ethernet device? Based on one of your earlier >> messages showing 26.4 Gbps between VMs on the same host, it seems as if >> this "limit" doesn't really matter. Unless I'm missing something? (I >> apologize if so.) >> >> >> > Thanks for your response! and I appreciate the help. >> > >> > - Ed >> > >> > >> > ps. Screen shots showing: >> > scenario 1: VM host b, VM host c on top, host a on bottom ( showing >> bond0 capping at 10, eth3 on host a taking 1/2 of the traffic ) >> > scenario 2: Host B, Host C on top. Host A on bot >> > >> > >> > >> > <scenario 1.png> >> > <scenario2.png> >> > >> > >> > >> >> On May 8, 2016, at 4:43 PM, Scott Lowe <scott.l...@scottlowe.org> >> wrote: >> >> >> >> Please see my response below. >> >> >> >> >> >>> On May 7, 2016, at 4:47 AM, ed bond <celpa.f...@gmail.com> wrote: >> >>> >> >>> Hello all, >> >>> >> >>> I was hoping someone might be able to help me diagnose what might be >> going on. >> >>> >> >>> Right now I have a bond0 interface setup with jumbo packets. I can >> get 18gigabit/s throughput to a single host. However inside the vms I am >> limited by 10gigabits per second. The VM’s have working jumbo packets. >> >>> >> >>> <snip> >> >>> >> >>> Any insights or open would be appreciated. >> >>> >> >>> Thanks! >> >> >> >> >> >> LACP bonds only help with aggregate throughput consisting of multiple >> streams of traffic between multiple endpoints. Any single stream (such as a >> VM talking to an endpoint on the network) is limited to the speed of one >> link within the bond. In this case, I'm guessing you have two 10Gbps links >> in the bond; therefore, a VM on the host talking to another endpoint on the >> network will be limited to 10Gbps. >> >> -- >> Scott >> >> > > _______________________________________________ > discuss mailing list > discuss@openvswitch.org > http://openvswitch.org/mailman/listinfo/discuss > >
_______________________________________________ discuss mailing list discuss@openvswitch.org http://openvswitch.org/mailman/listinfo/discuss
