All of lore.kernel.org
 help / color / mirror / Atom feed
* [LARTC] Best way to limit the rate
@ 2002-11-29 13:10 Mathieu Deziel
  2002-11-29 18:36 ` Stef Coene
  0 siblings, 1 reply; 2+ messages in thread
From: Mathieu Deziel @ 2002-11-29 13:10 UTC (permalink / raw)
  To: lartc

Suppose you need to limit the rate of the traffic that flows in an HTB
leaf class to 32kbits/sec

I can think of three ways to do it:

1.
make rate = ceil = 32kbit when creating this HTB leaf class.  A pfifo
qdisc can then be attached to this HTB leaf class to reduce the queue
size, and therefore reduce the latency of the packets.

2.
Attach a TBF qdisc to this HTB leaf class, and configure it properly.

3.
Have a policing filter before the HTB leaf class that will limit the
rate of the traffic before it enters the HTB leaf class.

Which one of these methods is the best?
My tests have shown that the latency of the packets is dramatically
better with method 3.  Anybody has an idea why this is the case?

--
Mathieu Deziel

_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: [LARTC] Best way to limit the rate
  2002-11-29 13:10 [LARTC] Best way to limit the rate Mathieu Deziel
@ 2002-11-29 18:36 ` Stef Coene
  0 siblings, 0 replies; 2+ messages in thread
From: Stef Coene @ 2002-11-29 18:36 UTC (permalink / raw)
  To: lartc

On Friday 29 November 2002 14:10, Mathieu Deziel wrote:
> Suppose you need to limit the rate of the traffic that flows in an HTB
> leaf class to 32kbits/sec
>
> I can think of three ways to do it:
>
> 1.
> make rate = ceil = 32kbit when creating this HTB leaf class.  A pfifo
> qdisc can then be attached to this HTB leaf class to reduce the queue
> size, and therefore reduce the latency of the packets.
>
> 2.
> Attach a TBF qdisc to this HTB leaf class, and configure it properly.
HTB _is_ TBF.  So the first idea is better.  A ceiled htb class = tbf qdisc.

> 3.
> Have a policing filter before the HTB leaf class that will limit the
> rate of the traffic before it enters the HTB leaf class.
>
> Which one of these methods is the best?
> My tests have shown that the latency of the packets is dramatically
> better with method 3.  Anybody has an idea why this is the case?
Solution 3 has the advantage that you limit the packet at a rate, but you 
don't have a queue.  The policer uses a small tbf queue to limit the packets.  
The packets are simply dropped if they exceed a certain rate.  It's also not 
possible to share bandwidth like you can do with cbq/htb.

Stef


-- 

stef.coene@docum.org
 "Using Linux as bandwidth manager"
     http://www.docum.org/
     #lartc @ irc.oftc.net

_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2002-11-29 18:36 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2002-11-29 13:10 [LARTC] Best way to limit the rate Mathieu Deziel
2002-11-29 18:36 ` Stef Coene

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.