[LARTC] Re: Need cbq or htb optimal solution

[Arman]

Hi,

    Please someone tell me what is the best in terms of performance when I
use one class per client or one class per package I am new to it and dont
know how the details. I am getting trouble sometimes when I connect from LAN
to server through SSH(putty), the connection starts to hang and speed goes
down. I dont know what configuration is wrong in cbq. Previously i was using
squid delay pools.

Regards,
Arman


On 5/26/07, Arman <arman.anwar at gmail.com> wrote:
>
> Hi all,
>
>
> Can anyone in this mailing list answer a few theoretical question which r
> confusing me.
>
> Here is the scenario
>
> I have a total Bandwidth of 2Mbps for a private LAN I am managing. I am
> using cbq standard script available online and for controlling bandwidth,
> squid and iptables. I have diff. packages for client.
>
> One: server own services access (unlimited bandwidth means no delays or
> control) + A Internet bandwidth
>
> Two: server own services (unlimited) + B Internet bandwidth
>
> Three: server own services (unlimited) + C Internet bandwidth
>
> Now I have to distribute speed between these clients, which are around
> 120. If I calculate I should have over 4Mbps total bandwidth. Currently I am
> using cbq for controlling bandwidth and have 120 cbq classes, one class per
> user. Is this good or I should have 4 categories of classes and users
> sharing the speed. Currently I figure out that in diff times of day
> bandwidth utilization is diff. sometimes users utilizing low bandwidth and
> most of bandwidth is wasted and sometimes scenario is diff.
>
> What will be the case if I switch to htb. I thought I could utilize ceil
> parameter to utilize wasted bandwidth in current configuration like one
> class per user.
>             Please suggest the optimal solution.
>
> Thnax is advance
> --
> Regards,
> M Arman




-- 
Regards,
Arman
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.ds9a.nl/pipermail/lartc/attachments/20070529/e457b0eb/attachment.htm