[ovs-discuss] OVS queues and precedence

Łukasz Rząsik lukasz.rzasik at gmail.com
Mon Nov 13 19:13:57 UTC 2017


2017-10-25 0:05 GMT+02:00 Ben Pfaff <blp at ovn.org>:

> On Wed, Oct 18, 2017 at 09:17:53AM +0200, Jannis Ohms wrote:
> > Hi
> >
> >
> > I used the following command to create 4 queues
> >
> > ovs-vsctl -- set port s7-eth2 qos=@newqos -- --id=@newqos create qos
> > type=linux-htb other-config:max-rate=10000000 queues:1=@newQ
> queues:2=@newQ2
> > queues:3=@newQ3 -- --id=@newQ create queue other-config:min-rate=2000000
> > other-config:max-rate=2000000,2=@newQ2 -- --id=@newQ2 create queue
> > other-config:min-rate=2000000 other-config:max-rate=5000000,3=@newQ3 --
> > --id=@newQ3 create queue other-config:min-rate=2000000
> > other-config:max-rate=6000000
> >
> > Now I wanted to know which queues precede each other i.e. if every queue
> has
> > a token and a packet to send who comes first. I requier this knowledge to
> > address the latency aspect of my QoS.
> >
> > I looked into Linux tc-htb manpage (https://linux.die.net/man/8/tc-htb)
> >
> > They define a prio property which seems to define  the precedence between
> > the queues.
> >
> > Can I pass this as an aditional parameter via ovs-vsctl
>
> OVS doesn't currently support the prio (priority) setting, and currently
> sets all priorities to 0.  If you want to submit a patch to add support
> for this feature, I think it would be a reasonable addition.
> _______________________________________________
> discuss mailing list
> discuss at openvswitch.org
> https://mail.openvswitch.org/mailman/listinfo/ovs-discuss
>

Hi Ben, Jannis,

I was about to implement prio (priority) setting for htb but I realized it
is actually implemented.
At least when I set other-config:priority for a given port then I can see
it set correctly with tc class show.

Am I right or am I missing something? Maybe you meant something else?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openvswitch.org/pipermail/ovs-discuss/attachments/20171113/4f3058ea/attachment-0001.html>


More information about the discuss mailing list