Re: [PATCH net-next 4/4] net: Add Open vSwitch kernel components.

[jamal <hadi-fAAogVwAN2Kw5LPnMra/2Q@public.gmane.org>]

On Sat, 2011-11-26 at 23:17 -0800, Justin Pettit wrote:

> 
> You're right--calling tc directly through system() is kind of ugly.
> That code was written a *long* time ago when we wanted a quick QoS story.
> As you mentioned, we use netlink to configure traffic shaping, so we have
> all the pieces at this point.  I just think no one ever bothered to clean
> up that little wart in userspace.  

Understandable.  You shipped something that works.

I guess the point i was trying to make is you _are_ already using
the classifier action subsystem. You have the u32 classifier and
the policer action. You should just introduce a new classifier
and the 2 or so actions you need and other folks can benefit from
them as well without need to use openvswitch.

> We encourage users to use shaping, since it generally provides 
> better results (and we do expose per-flow granularity there).  
> As a result, we haven't seen a need to improve support for policing.

HTB's metering algorithm was essentially originally ripped off the
policer action; probably better off for TCP to use shaping hence the
results you are observing.

cheers,
jamal