I actually asked them: http://lists.opencontrail.org/pipermail/users_lists.opencontrail.org/2014-July/000338.html
As I understood it, they are able to offload from VM to VM, without segmentation - maybe I’m wroing, and what actually is happening is that their vrouter module, on the receiving side, is able to reassemble the packets to 64k before sending to the vif. This reassembly I cannot reproduce with OVS. Maybe this is how they reach the 9.18Gbps ? On 29/07/2014, at 03.56.51, Han Zhou <zhou...@gmail.com> wrote: > Hi, >> >> OK - I see, thanks for clearing that up. >> Do you know if VXLAN has the same limitation ? >> The Contrail vrouter seems to be able to work around it.. >> http://opencontrail.org/evaluating-opencontrail-virtual-router-performance/ >> > > Yes, VXLAN is the same. I just read the OpenContrail link you > mentioned. It seems impossible to me to achieve 9.18Gbps for MTU 1500 > without hardware offloading... We did some work on VXLAN based on OVS, > and the max throughput we could achieve was about 6.7Gbps, before > optimization it was 2.6Gbps. > > Would be interested to study their code. > > Best regards, > Han > _______________________________________________ > discuss mailing list > discuss@openvswitch.org > http://openvswitch.org/mailman/listinfo/discuss > _______________________________________________ discuss mailing list discuss@openvswitch.org http://openvswitch.org/mailman/listinfo/discuss