On Fri, 21 Sep 2018 23:01:12 +0200
Damjan Marion <dmar...@me.com> wrote:

> > On 21 Sep 2018, at 20:27, Stephen Hemminger <step...@networkplumber.org> 
> > wrote:
> > 
> > On Thu, 20 Sep 2018 21:04:17 +0200
> > Damjan Marion <dmar...@me.com <mailto:dmar...@me.com>> wrote:
> >   
> >>> On 20 Sep 2018, at 20:27, Stephen Hemminger <step...@networkplumber.org> 
> >>> wrote:
> >>> 
> >>> Is there a way to use virtio_user as an exception path with DPDK?
> >>> 
> >>> https://doc.dpdk.org/guides-17.11/howto/virtio_user_as_exceptional_path.html
> >>> 
> >>> Testing simple case, I see much better performance with this when 
> >>> forwarding
> >>> a 10G device to kernel.
> >>> 
> >>>           Rx              Tx
> >>>   KNI     3.2 Gbit/sec    1.3 Gbit/sec
> >>>   TAP     4.9             4.7
> >>>   Virtio  5.6             8.6
> >>> 
> >>> Maybe a VPP plugin is needed.    
> >> 
> >> Isn't that what we have in our new TAP code? TAP with vhost-net backend?
> >> 
> >> https://git.fd.io/vpp/tree/src/vnet/devices/tap/tap.c 
> >> <https://git.fd.io/vpp/tree/src/vnet/devices/tap/tap.c> 
> >> <https://git.fd.io/vpp/tree/src/vnet/devices/tap/tap.c 
> >> <https://git.fd.io/vpp/tree/src/vnet/devices/tap/tap.c>>  
> > 
> > The performance for simple passthrough/bridge of 10 Gig NIC into TAP seems
> > to be quite slow with VPP and has lots of bufferbloat.
> > 
> > See attached rrul test results:
> >                             Down            Up              Latency
> >     DPDK TAP                550 Mbit/sec    700 Mbit/sec    3.5ms
> >     DPDK virtio_user        800             800             2.5
> >     VPP TAP                  50             1000            10
> >   
> 
> Hard to comment as no much date about your config. I never looked into dpdk 
> code, might be that they do something very special :)
> 
> 
You can also refer to this paper for more details:
https://dl.acm.org/citation.cfm?id=3098586

You can get a virtio_user device in VPP, via dpdk vdev but it doesn't seem to 
work the same as tap.
L2 bridge isn't working as expected. Maybe promiscuous doesn't work.

dpdk {
        socket-mem 1024,0
        vdev virtio_user0,path=/dev/vhost-net,queue_size=1024
}

vpp# show int
              Name               Idx    State  MTU (L3/IP4/IP6/MPLS)     
Counter          Count     
GigabitEthernet8/0/0              2     down         9000/0/0/0     
TenGigabitEthernet4/0/0           1     down         9000/0/0/0     
VirtioUser0/0/0                   3     down         9000/0/0/0     
local0                            0     down          0/0/0/0   
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.

View/Reply Online (#10605): https://lists.fd.io/g/vpp-dev/message/10605
Mute This Topic: https://lists.fd.io/mt/25834626/21656
Group Owner: vpp-dev+ow...@lists.fd.io
Unsubscribe: https://lists.fd.io/g/vpp-dev/unsub  [arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-

Reply via email to