[ovs-discuss] QoS Egress Port Traffic Shaping (linux-hfsc ) not Working as Expected for UDP Traffic

[Ramzah Rehman]

Thanks, Ben. I figured out that even without apply any QoS or queueing, UDP
traffic observes throughout of around 175Mbps. I guess the issue is not
with QoS or queueing.


On Sat, Dec 15, 2018, 1:08 AM Ben Pfaff <blp at ovn.org wrote:

> On Fri, Dec 14, 2018 at 05:53:05PM +0500, Ramzah Rehman wrote:
> >  I have two servers connected via 1Gbps cable. I have installed OVS 2.8
> on
> > server one. I have a switch named OVS_BR_LEAF_1 on it. I experimented
> with
> > QoS traffic shaping. I have connected eth3 of server one to ovs-switch on
> > port 2. I want traffic going from server one to server two to observer
> QoS
> > traffic rate that I specify.
> >
> > I added following configuration:
> >
> > #ovs-vsctl -- set port eth3 qos=@newqos -- --id=@newqos create qos
> > type=linux-hfsc other-config:max-rate=1000000000 queues:1=@q1 -- --id=@q1
> > create queue other-config:min-rate=x other-config:max-rate=x
> >
> > I have following flow entries in my switch:
> > #vs-ofctl add-flow OVS_BR_LEAF_1
> > priority=6000,in_port=LOCAL,actions=set_queue:1,normal
> > #ovs-ofctl add-flow OVS_BR_LEAF_1 priority=6000,in_port=2,actions=normal
> >
> > Then I checked throughput for TCP traffic (from server one to two) via
> > iperf and got following results:
> >
> > got 846 Mbps for x=900 Mbps
> > got 757 Mbps for x=800 Mbps
> > got 653 Mbps for x=700 Mbps
> > got 428 Mbps for x=450 Mbps
> > got 381 Mbps for x=400 Mbps
> > got 287 Mbps for x=300 Mbps
> > got 239 Mbps for x=250 Mbps
> > got 192 Mbps for x=200 Mbps
> >
> > Seems like traffic shaping is working fine for TCP traffic. However, for
> > UDP, I got these results:
> >
> > 148 Mbps for x=800 Mbps
> > 185 Mbps for x=300 Mbps
> > 131 Mbps for x=200 Mbps
> > 93 Mbps for x=100 Mbps
> > 77 Mbps for x=80 Mbps
> > 28.9 Mbps for x=30 Mbps
> > 19.3 Mbps for x=20 Mbps
> >
> > For UDP, traffic shaping works fine till x=100Mpbs, but for higher values
> > is shows unexpected behavior. Is there an explanation?
>
> This FAQ entry probably applies.
>
> Q: I configured QoS, correctly, but my measurements show that it isn't
> working
> as well as I expect.
>
>     A: With the Linux kernel, the Open vSwitch implementation of QoS has
> two
>     aspects:
>
>     - Open vSwitch configures a subset of Linux kernel QoS features,
> according
>       to what is in OVSDB.  It is possible that this code has bugs.  If you
>       believe that this is so, then you can configure the Linux traffic
> control
>       (QoS) stack directly with the "tc" program.  If you get better
> results
>       that way, you can send a detailed bug report to bugs at openvswitch.org
> .
>
>       It is certain that Open vSwitch cannot configure every Linux kernel
> QoS
>       feature.  If you need some feature that OVS cannot configure, then
> you
>       can also use "tc" directly (or add that feature to OVS).
>
>     - The Open vSwitch implementation of OpenFlow allows flows to be
> directed
>       to particular queues.  This is pretty simple and unlikely to have
> serious
>       bugs at this point.
>
>     However, most problems with QoS on Linux are not bugs in Open vSwitch
> at
>     all.  They tend to be either configuration errors (please see the
> earlier
>     questions in this section) or issues with the traffic control (QoS)
> stack
>     in Linux.  The Open vSwitch developers are not experts on Linux traffic
>     control.  We suggest that, if you believe you are encountering a
> problem
>     with Linux traffic control, that you consult the tc manpages (e.g.
> tc(8),
>     tc-htb(8), tc-hfsc(8)), web resources (e.g. http://lartc.org/), or
> mailing
>     lists (e.g. http://vger.kernel.org/vger-lists.html#netdev).
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openvswitch.org/pipermail/ovs-discuss/attachments/20181215/04f125b1/attachment.html>