[ovs-discuss] network bandwidth in Openstack when using OVS+VLAN

Luca Giraudo lgiraudo at nicira.com
Tue Jul 30 05:36:50 UTC 2013


Few more things to check:

* are you using KVM? If yes, can you verify that your VMs are using
vhost_net, please? In the KVM process related to the VM there should be a
vhost=on parameter. If not, modprobe vhost_net.

* I saw different throughput with different OS in the VM. Did you try
different VMs?

* what bandwidth do you get among two VMs in the same compute node?

* can you monitor the CPU usage in the compute node and look out for big
CPU consumer, please?

Thanks,
Luca
On Jul 29, 2013 6:26 PM, "Li, Chen" <chen.li at intel.com> wrote:

>  ** Is the VM ethernet driver a para-virtual driver? Para-virtual drivers
> give a good performance boost.*
>
> I have used openstack default parameters, it is virtio, I think virtio
> should have a good performance:****
>
>     <interface type='bridge'>****
>
>       <mac address='fa:16:3e:ca:4a:86'/>****
>
>       <source bridge='br-int'/>****
>
>       <virtualport type='openvswitch'>****
>
>         <parameters interfaceid='3213dbec-f2ea-462f-818b-e07b76a1752c'/>**
> **
>
>       </virtualport>****
>
>       <target dev='tap3213dbec-f2'/>****
>
>       <model type='virtio'/>****
>
>       <address type='pci' domain='0x0000' bus='0x00' slot='0x03'
> function='0x0'/>****
>
>     </interface>****
>
> ** **
>
> ** **
>
> ** Is TSO ON in the VM and the Hypervisor?*
>
> * *
>
> The VM:****
>
> Features for eth0:****
>
> rx-checksumming: off [fixed]****
>
> tx-checksumming: on****
>
>         tx-checksum-ipv4: off [fixed]****
>
>         tx-checksum-ip-generic: on****
>
>         tx-checksum-ipv6: off [fixed]****
>
>         tx-checksum-fcoe-crc: off [fixed]****
>
>         tx-checksum-sctp: off [fixed]****
>
> scatter-gather: on****
>
>         tx-scatter-gather: on****
>
>         tx-scatter-gather-fraglist: on****
>
> *tcp-segmentation-offload: on*
>
>         tx-tcp-segmentation: on****
>
>         tx-tcp-ecn-segmentation: on****
>
>         tx-tcp6-segmentation: on****
>
> udp-fragmentation-offload: on****
>
> generic-segmentation-offload: on****
>
> generic-receive-offload: on****
>
> large-receive-offload: off [fixed]****
>
> rx-vlan-offload: off [fixed]****
>
> tx-vlan-offload: off [fixed]****
>
> ntuple-filters: off [fixed]****
>
> receive-hashing: off [fixed]****
>
> highdma: on [fixed]****
>
> rx-vlan-filter: on [fixed]****
>
> vlan-challenged: off [fixed]****
>
> tx-lockless: off [fixed]****
>
> netns-local: off [fixed]****
>
> tx-gso-robust: off [fixed]****
>
> tx-fcoe-segmentation: off [fixed]****
>
> fcoe-mtu: off [fixed]****
>
> tx-nocache-copy: on****
>
> loopback: off [fixed]****
>
> rx-fcs: off [fixed]****
>
> rx-all: off [fixed]****
>
> The hypervisor:****
>
> Offload parameters for eth4:****
>
> rx-checksumming: on****
>
> tx-checksumming: on****
>
> scatter-gather: on****
>
> *tcp-segmentation-offload: on*
>
> udp-fragmentation-offload: off****
>
> generic-segmentation-offload: on****
>
> generic-receive-offload: on****
>
> large-receive-offload: off****
>
> rx-vlan-offload: on****
>
> tx-vlan-offload: on****
>
> ntuple-filters: off****
>
> receive-hashing: on****
>
> ** **
>
> ** What throughput do you get while using Linux bridge instead of OVS?*
>
> Currently, I don’t have a linux bridge environment.****
>
> But, I remember in virtio test, while I create bridge by hand and assigned
> it to an instance, I can always get near hardware limitation bandwidth if I
> have enough threads.****
>
> ** **
>
> ** Are you using tunnels? If you are using a tunnel like GRE, you will
> see a throughput drop.*
>
> No, I’m working under Quantum+OVS+*VLAN*.****
>
> ** **
>
> Thanks.****
>
> -chen****
>
> ** **
>
> *From:* Gurucharan Shetty [mailto:shettyg at nicira.com]
> *Sent:* Tuesday, July 30, 2013 12:06 AM
> *To:* Li, Chen
> *Cc:* discuss at openvswitch.org
> *Subject:* Re: [ovs-discuss] network bandwidth in Openstack when using
> OVS+VLAN****
>
> ** **
>
> There could be multiple reasons for the low throughput. I would probably
> look at the following.****
>
> ** **
>
> * Is the VM ethernet driver a para-virtual driver? Para-virtual drivers
> give a good performance boost.****
>
> * Is TSO ON in the VM and the Hypervisor?****
>
> * What throughput do you get while using Linux bridge instead of OVS?****
>
> * Are you using tunnels? If you are using a tunnel like GRE, you will see
> a throughput drop.****
>
> ** **
>
> ** **
>
> On Mon, Jul 29, 2013 at 1:48 AM, Li, Chen <chen.li at intel.com> wrote:****
>
> Hi list,****
>
>  ****
>
> I’m a new user to OVS.****
>
>  ****
>
> I installed OpenStack Grizzly, and  using Quantum + OVS + VLAN for network.
> ****
>
>  ****
>
> I have two compute nodes with 10 Gb NICs, and the bandwidth between them
> is about  *8.49* Gbits/sec (tested by iperf).****
>
>  ****
>
> I started one instance at each compute node:****
>
> instance-a => compute1****
>
> instance-b=> compute2****
>
> The bandwidth between this two virtual machine is only *1.18* Gbits/sec.**
> **
>
>   ****
>
> Then I start 6 instances at each compute node:****
>
>           (   instance-a => compute1 ) ----- iperf------ > (instance-b=>
> compute2)****
>
>                           (   instance-c => compute1 ) ----- iperf------ >
> (instance-d=> compute2)****
>
>                           (   instance-e => compute1 ) ----- iperf------ >
> (instance-f=> compute2)****
>
>                           (   instance-g => compute1 ) ----- iperf------ >
> (instance-h=> compute2)****
>
>                           (   instance-i => compute1 ) ----- iperf------ >
> (instance-j=> compute2)****
>
>                           (   instance-k => compute1 ) ----- iperf------ >
> (instance-l=> compute2)****
>
> The total bandwidth is only *4.25 *Gbits/sec.****
>
>  ****
>
>  ****
>
> Anyone know why the performance is this low ?****
>
>  ****
>
> Thanks.****
>
> -chen****
>
>
> _______________________________________________
> discuss mailing list
> discuss at openvswitch.org
> http://openvswitch.org/mailman/listinfo/discuss****
>
>  ** **
>
> _______________________________________________
> 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/20130729/5e53e239/attachment.html>


More information about the discuss mailing list