In article <s63011-(E-Mail Removed)>, Andy Furniss wrote:
>
>> I would appreicate feedback on it though.
>
> I'll give it a try - not that I've got much of a network.
>
> I tried a while ago to patch my iptables 1.28 with the IMQ POM - but it was
> for the CVS & didn't work by just changing dir names.
>
> Do the normal patches against the older iptables just work with 1.28?
>
I have been running the 2.4.18 (old) patches with the 2.4.21 kernel until
today with no problem and on lots of other peoples machine. However today I
have updated the script and also changed everything to use the new 2.4.21
patches from the IMQ website; so you might as well use the new stuff
> This script is really only suitable for situations where you want to fairly
> share and prioritize the traffic for a group of machines on a per IP basis
>
> Do you mean it will give me 50% of the down bandwidth if I had one tcp
> connection, while another user had 10 on the go - assuming they are the
> same sort of traffic.
>
With all the other QoS scripts I have seen out there they fair share on a
'flow' (TCP/UDP/whatever, per connection) basis, which is okay as on average
everyone over a long period of time will use a similar number of 'flows'.
However in what if one computer has ten 'flows' (say ten ftp download
connections) and another person has only two, then the bandwidth is split 5:1
which obviously is quite unfair. This can be gotten round in a kludgy way
with a script that makes a tree of buckets for each machine, my QoS script
uses the ESFQ system though which fair shares on a per 'ip' basis instead.
Therefore in the above example still only 1:1 sharing occurs, which is great.
The disadvantage is that you need a kernel patch and a 'tc' patch, however
its not the end of the world and well worth the effort.
I hope this clears things up?
You might want to be lazy and just get my 'jdg' kernel patch
(
http://digriz.org.uk/jdg-kernel-patch/)
Regards
Alex
--
Alexander Clouter <(E-Mail Removed)>
MetroNet Support
http://www.metronet.co.uk/support/