All of lore.kernel.org
 help / color / mirror / Atom feed
From: Andy Furniss <andy.furniss@dsl.pipex.com>
To: lartc@vger.kernel.org
Subject: Re: [LARTC] strange behaviour of qos
Date: Sat, 19 Mar 2005 11:53:37 +0000	[thread overview]
Message-ID: <423C12C1.4050408@dsl.pipex.com> (raw)
In-Reply-To: <42319DC2.80500@post.pl>

Ratel wrote:
> Ed Wildgoose wrote:
> 
>>
>>>
>>> any idea what might be wrong ?
>>
>>
>>
>>
>> Yep, something is queuing somewhere...
>>
>> Either your outbound setup allows too much outbound for the outbound 
>> connection (and if you use ADSL remember packets use more bandwidth 
>> than on ethernet).  Or you aren't throttling your inbound connection 
>> enough and queuing is occuring on the ISP end of your link
>>
>> Ed W
>>
> No, I really don't think so. The link itself is a 100%-CIR SDSL (for 
> LANof 300+ machines). I've
> set up ceil(up and down) to ~95% of real bandwidth.

That may be OK as long as CIR is at ip level not atm/aal5 also if you 
use older tc k means 1024.

In the case if ingress traffic - even if you set 95% and it really is 
95% then you need to back off a bit as the traffic has already been 
shaped by the link and at 95% you are only going to queue 5 packets for 
every hundred that pass - which means tcp bursts can fill and maybe drop 
from the buffer at the other end - hurting latency and if dropping - 
your bandwidth allocation.

> It seems that classes exceed maximum rates (ceils) defined for them.
> For example : p2p class has ceil of 2.5 mbits
> When p2p traffic is not allowed to pass through bandwidth usage varies 
> around 30%.
> Afrer allowing p2p to pass bandwidth usage quickly rises to nearly 100%.
> Is it possible to somehow isolate a class (as in cbq?) ?. In the 
> archives I've
> found something like this (let's say): '[...] rate 2500kbit ceil 
> 2500kbit'. This should
> produce something like isolated class , but for me it simply doesn't work.
> I'd be very grateful for any suggestions / ideas ...
> 
> Ratel
> 
> _______________________________________________
> LARTC mailing list
> LARTC@mailman.ds9a.nl
> http://mailman.ds9a.nl/cgi-bin/mailman/listinfo/lartc
> 

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

      parent reply	other threads:[~2005-03-19 11:53 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2005-03-11 13:31 [LARTC] strange behaviour of qos Ratel
2005-03-11 13:50 ` Ed Wildgoose
2005-03-12 16:58 ` Ratel
2005-03-19 11:53 ` 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=423C12C1.4050408@dsl.pipex.com \
    --to=andy.furniss@dsl.pipex.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.