All of lore.kernel.org
 help / color / mirror / Atom feed
From: Andy Furniss <lists@andyfurniss.entadsl.com>
To: lartc@vger.kernel.org
Subject: Re: [LARTC] Fair shaping over link with variable parameters
Date: Thu, 01 Jun 2006 20:24:38 +0000	[thread overview]
Message-ID: <447F4D06.30607@andyfurniss.entadsl.com> (raw)
In-Reply-To: <20060528193129.GA15843@iceberg.netwerke.eu.org>

Rafal Krypa wrote:
> Hi.
> I would like to ask you for advice.
> I am trying to construct following shaping solution:
> * several users are using one link to the Internet
> * all of them have equal priority and should be given fair amount of bandwidth
> * no kind of traffic is considered more important than other
> * our Internet connection has no CIR, only "maximum dl/ul speeds" given by
>   provider

What you can or can't do will depend on the exact nature and behavior of 
the link.

> * most important: our outgoing and incoming traffic must be shaped to some rate
>   that will provide possibly low latency. For users that do not have active
>   connections I'd like to ensure no more than 100ms latency for ping or any
>   other low-traffic connections

100ms - that would be hard to guarantee on a slow fixed rate link, in 
some situations you may need to sacrifice 50% of ingress bandwidth.

It depends on how fast the link is and how slow it gets and how it is 
slowed.

>   
> 
> For several years of my experiments with traffic shaping over Linux I found no 
> tool for creating such system. For example, HTB require given, constant 'ceil' 
> parameter. I would like to have some qdisc that can automatically adjush its 
> rate/ceil parameter depending on achieved latency. The rest of the job would be 
> quite pretty done by ESFQ.
> Could you point me to anything adequate to my needs?
> 

There is no qdisc that has variable rates.

I've just got a link with variable down speed and have played around 
with policers to see what's possible. I haven't done much and it doesn't 
work too well - though it works enough to carry on trying to see what's 
possible. I still don't know whether it can ever work enough to be left 
"unnatended".

If you have few users and know your traffic and have a fairly fast link 
and know how it behaves there may be a way - at least to do alot better 
than doing nothing.

Andy.
_______________________________________________
LARTC mailing list
LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc

      parent reply	other threads:[~2006-06-01 20:24 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2006-05-28 19:31 [LARTC] Fair shaping over link with variable parameters Rafal Krypa
2006-05-29 13:00 ` Andreas Klauer
2006-05-29 14:31 ` Rafal Krypa
2006-06-01 20:24 ` Andy Furniss [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=447F4D06.30607@andyfurniss.entadsl.com \
    --to=lists@andyfurniss.entadsl.com \
    --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.