All of lore.kernel.org
 help / color / mirror / Atom feed
From: Stef Coene <stef.coene@docum.org>
To: lartc@vger.kernel.org
Subject: Re: [LARTC] Some questions remaining about TC
Date: Wed, 11 Jun 2003 16:47:53 +0000	[thread overview]
Message-ID: <marc-lartc-105535012524893@msgid-missing> (raw)
In-Reply-To: <marc-lartc-105519389407466@msgid-missing>

On Wednesday 11 June 2003 12:26, Emmanuel SIMON wrote:
> First, thank you Stef for answering.
> That's what I understood. I think my question was not enough precise (??)
>
> My reel question/problem is why are parameters of the bandwidth (such as
> rate, burts ...) set on classes rather than on qdiscs ?
You have to see a class as a virtual channel.  Each class is responsible for 
the packets that it holds.  So the rate / burst / ceil and so one are only 
valid for that class.  So you have to put packets in the same class that 
somehow are the same.  Like all telnet/ssh packets.  So you can give that 
class a small burst and a low latency.  Or you put the packets from 1 client 
in 1 class so all the traffic to the client is one 1 big channel and he is 
reponsible for this.

> I guess a class should only map a flow and should not know how this flow
> will be sent. Especially that classes don't send packets. What is the
> reason why it is done so ? And what if i don't specify parameters fir a
> leaf class ? Does it herit from the parent ?
A class is not reponsible for sending the packets because the class is only 
used to create the virtual channels (and so to limit the packets in that 
class).  Once a packet is dequeued from a virtual channel (and so is dequeued 
form a class) it's give to the qdisc attached to that class.  So you can add 
a small fifo qdisc to that class so packets are send fast.  Or you can add a 
sfq qdisc so each flow has his own little queue.  So you have a lot of 
flexibility.

You can even add a classfull qdisc to a class, but that's just a waste of CPU 
cycles and will only add extra delays.

Each class needs it's own parameters when you create it.  It will not inheret 
something from its parent.

> Thanks once more
This is even for more confusing when I have to think about ut.  This is in no 
docs explained, but I don't think that I'm telling big lies.

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/

      parent reply	other threads:[~2003-06-11 16:47 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2003-06-09 21:18 [LARTC] Some questions remaining about TC Emmanuel SIMON
2003-06-10 18:55 ` Stef Coene
2003-06-11  7:41 ` Lars Landmark
2003-06-11 10:26 ` Emmanuel SIMON
2003-06-11 16:47 ` Stef Coene [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=marc-lartc-105535012524893@msgid-missing \
    --to=stef.coene@docum.org \
    --cc=lartc@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
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.