Xen project Mailing List

On 23/07/2014, at 17.17.02, Kristoffer Egefelt <kristoffer@xxxxxxx> wrote:

Hi,

Right, traffic not leaving the dom0 seems not to be affected by this - but in this case:

- Two dom0s connected with 10G
- One VM (HVM/PV does not matter) on each dom0
- One vif in each VM

Unidirectional traffic between the two VMs reaches near line rate on mtu 1500 ~800.000 packets pr second.

I tried to tune the netback / dom0 alot, but I never saw a vif doing more than 200.000 pps … ?

Is virtio really that much faster - or am I doing something very wrong with XEN ?

On 23/07/2014, at 16.41.27, Eugene Istomin <e.istomin@xxxxxxx> wrote:

Hello,

>no reason I know of why Xen also couldn't do it.
We have 12-16 Gb/s for inter-Dom0 VMs, there are no slow network issues related to Xen 4.4 in case of inter-Dom0 & MTU 9000.

---
Best regards,
Eugene Istomin

On Wednesday, July 23, 2014 08:51:35 AM Steven Timm wrote:
> I have seen Xen do more than 1.5GBit/s on vm-to-vm performance
> within the same hypervisor. Never tried it on a machine with
> 10Gbit/s hardware. KVM is certainly capable of 10Gbit/s wire speed
> with virtio and we do it regularly, no reason I know of why Xen also
> couldn't do it.
>
> Steve
>
> On Wed, 23 Jul 2014, Kristoffer Egefelt wrote:
> > Hi,
> >
> > Until now I was under the impression that a VM could not move more than
> > ~200.000pps or ~1.5Gbit on mtu 1500 on one VIF - in XENs case due to the
> > netback process being single threaded.
> >
> > But here they apparently get line-rate 10G on a single vif one-core VM
> > (+800.000pps), using standard KVM and virtio (this is not documented in
> > the post, I asked them).
> >
> > http://opencontrail.org/evaluating-opencontrail-virtual-router-performance
> > /
> >
> > This seems almost unreal to me - but maybe I’m missing something.
> >
> > Is Xen currently able to do this ?
> >
> >
> > Thanks
> > Kristoffer
> > _______________________________________________
> > Xen-users mailing list
> > Xen-users@xxxxxxxxxxxxx
> > http://lists.xen.org/xen-users
>
> ------------------------------------------------------------------
> Steven C. Timm, Ph.D (630) 840-8525
> timm@xxxxxxxx http://home.fnal.gov/~timm/
> Fermilab Scientific Computing Division, Scientific Computing Services Quad.
> Grid and Cloud Services Dept., Associate Dept. Head for Cloud Computing

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxx
http://lists.xen.org/xen-users

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxx
http://lists.xen.org/xen-users

Re: [Xen-users] netback performance