All of lore.kernel.org
 help / color / mirror / Atom feed
From: Grant Taylor <gtaylor@riverviewtech.net>
To: lartc@vger.kernel.org
Subject: Re: [LARTC] load balancing
Date: Wed, 13 Dec 2006 22:42:03 +0000	[thread overview]
Message-ID: <458081BB.7020909@riverviewtech.net> (raw)
In-Reply-To: <marc-lartc-102512835209719@msgid-missing>

Charlie Meyer wrote:
> I ive set up a working linux router with load balancing as per the lartc 
> guide. Everything is working properly, except for the load balancing 
> does not seem to be equally balance the load. The first line gets about 
> 60% of the incoming load, the 2^nd line gets about 40%, and the 3^rd 
> line hardly gets any of it at all. The outgoing load is evenly balanced 
> among all three lines.
> 
> I did not set any weights when I set up the route, so this is confusing 
> to me. BTW, I have been using ntop as well as ibmonitor to view the load 
> over each line.
> 
> when I have tried setting weights, such as 1 for the first line, 2 for 
> the second line, and 4 for the second line, it still doesn’t balance 
> correctly.
> 
> I am using fedora core 6 with all the latest packages and kernel as per yum.
> 
> any ideas here would be greatly appreciated

Keep in mind that you do not have any control of the traffic that is 
inbound to you.  The only thing that you can control is the traffic that 
you send.

So, what is probably happening is that your system is load balancing the 
out bound traffic, which is being replied to by the server.  Consider 
you have 3 connections, one out bound request each.  Unless you are 
dealing with globally routable IP addresses behind your system and are 
not NATing at your system, responses to out bound requests will come 
back in to the same IP that the request originated from.  With that in 
mind, consider one reply being a 512 byte response, one reply being a 1k 
byte response, and one reply being a 20 k byte response.  In this case, 
one connection will receive 512 bytes, another 1 k byte, and the last 20 
k bytes.

I think we can all agree that this is FAR from a load balanced scenario. 
  However, we do not have any control over the reverse route, this is at 
best our providers control.  If each link is with a different provider, 
there is no way to load balance the traffic back to our system.  If, by 
chance the links are all with one provider and they are willing to work 
with you and you do not have any reverse path filtering in place, the 
provider could spread the load across all the links evenly.  However, 
this is way beyond the scope of "Load Balancing" under Linux, or any 
thing else for that matter, and thus is more or less just accepted.

If you would like, I can go in to more depth as to why this does not 
work as is and what would have to be done to make this work.

Incidentally, this is also why QoS does not really work well on inbound.



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

      parent reply	other threads:[~2006-12-13 22:42 UTC|newest]

Thread overview: 23+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2002-06-26 21:52 [LARTC] Load balancing Greg Scott
2002-06-28  5:27 ` William L. Thomson Jr.
2002-07-04 14:52 ` bert hubert
2003-02-19 19:27 ` Paul English
2003-05-08 11:50 ` [LARTC] load balancing Lucas Aimaretto
2003-05-08 13:08 ` Sjaak Nabuurs
2003-05-09 15:44 ` Víctor Nuño
2003-05-09 15:55 ` Martin A. Brown
2003-11-28  1:41 ` Евгени Гечев
2003-11-28 12:38 ` Ivo Vachkov
2003-11-28 18:47 ` Evgeni Gechev
2003-11-28 19:02 ` Artūras Šlajus
2003-11-28 23:22 ` Ivo Vachkov
2004-07-26  1:35 ` [LARTC] Load Balancing XMundo - Soporte Tecnico
2004-07-26 23:58 ` XMundo - Soporte Tecnico
2004-09-15 17:10 ` [LARTC] load balancing pbruna
2004-09-17  3:21 ` Glen Mabey
2005-01-12 21:15 ` Antonio Pérez
2005-11-16 20:13 ` comp.techs
2006-10-13  8:54 ` Indunil Jayasooriya
2006-10-13  8:55 ` Indunil Jayasooriya
2006-12-13 22:07 ` Charlie Meyer
2006-12-13 22:42 ` Grant Taylor [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=458081BB.7020909@riverviewtech.net \
    --to=gtaylor@riverviewtech.net \
    --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.