<div>On Tue, Jan 12, 2010 at 4:02 PM, <span dir="ltr"><<a href="mailto:clauz@ninux.org">clauz@ninux.org</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
<div><div class="h5">On 01/12/2010 03:10 PM, Michael Rack wrote:<br>
><br>
>> ... to specify the available bandwith. Otherwise (i.e. with "infinite"<br>
>> bandwidth available") the tc queues will always have zero packets and<br>
>> prioritization will not work.<br>
><br>
> Markus Kittenberger has posted a short snipped of the QOS Setup.<br>
> Message-ID: <a href="mailto:op.u6etk32h871hnn@markit.mshome.net">op.u6etk32h871hnn@markit.mshome.net</a><br>
><br>
> This setup handle 3 SFQs (Stochastic Fairness Queuing). The filter<br>
> methods have priority settings 1,2,3,4 and 5. With this setup you don't<br>
> have to know the uplink speed. Matching packets that are queued will<br>
> reordered as defined the rest is done by your OS network-stack.<br>
<br>
<br>
</div></div>The man page of SFQ [<a href="http://linux.die.net/man/8/tc-sfq" target="_blank">http://linux.die.net/man/8/tc-sfq</a>] says:<br>
<br>
"""<br>
Please note that SFQ, like all non-shaping (work-conserving) qdiscs, is<br>
only useful if it owns the queue. This is the case when the link speed<br>
equals the actually available bandwidth. This holds for regular phone<br>
modems, ISDN connections and direct non-switched ethernet links.<br>
<br>
Most often, cable modems and DSL devices do not fall into this category.<br></blockquote><div>because they are potentially doing their own packet reordering and droppping, and have usually a faster link to your router than to the other end of the DSL or whatever,..</div>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
The same holds for when connected to a switch and trying to send data to<br>
a congested segment also connected to the switch.<br>
<br>
In this case, the effective queue does not reside within Linux and is<br>
therefore not available for scheduling.<br>
<br>
Embed SFQ in a classful qdisc to make sure it owns the queue.<br></blockquote><div>fyi The PRIO qdisc is a classful qdisc! </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
"""<br>
<br>
I think that:<br>
1) Wireless links do not fall into the category of "link speed being<br>
equal to the available bandwidth".<br></blockquote><div><br></div><div><div>1 is true imho<br></div><div><br></div><div>noone says that the speed must be static *G</div> </div></div>