[LARTC] Shape own router

[Salatiel Filho]

On 3/24/07, Andy Furniss <lists at andyfurniss.entadsl.com> wrote:
>
> Salatiel Filho wrote:
>
> >
> > Hi Andy , thanks again , but i am not understanding very well how to
> > do it [still newbie in this].  Let`s try to change to some real code
> > here. This is part of my  setup to shape download:
> >
> > eth0 = EXTIF
> > eth1 = LOCALIF
> >
> > # SHAPE DOWNLOAD to LOCALNET NOT COMING FROM THE ROUTER ITSELF [samba
> > for example]
> > iptables -t mangle -s ! 192.168.254.254 -A POSTROUTING -o eth1 -j IMQ
> > --todev 1
>
> If you shape your wan - eth0 using ifb on ingress or imq from prerouting
> then you do not need any rules on eth1, the wan traffic will already be
> shaped.
>
> If you do not plan on seperating users or interactive traffic from bulk
> traffic, it would actually be much nicer to use a policer for ingress
> wan traffic. Policing doesn't buffer traffic just drops it when a
> virtual buffer is full, so you won't be delaying interactive traffic by
> queuing with bulk.
>
> When you shape ingress wan, however you do it, you will need to
> sacrifice about 20% of your bandwidth, possibly more depending on
> needs/traffic/wan speed. Shaping from the wrong end of the bottleneck is
> better than doing nothing, but you can't do it perfectly.
>
> >
> > tc qdisc add dev imq1 root handle 1: htb  default 3 r2q 1   //
> > DOWNLOAD SHAPER ROOT
> > tc class add dev imq1 parent 1: classid 1:1 htb rate 2048kbit quantum
> > 1500 //KNOWN TRAFFIC GOES HERE
> > tc class add dev imq1 parent 1: classid 1:3 htb rate 8kbit quantum
> > 1500  // DEFAULT CLASS VERYYYYY SLOWWWWWWW
>
> If this were eth rather than imq you would be sending arp to a slow
> class - not nice.
>
> Andy.
>
>
I was finally able to shape the router itself :) , i changed IMQ default
behaviour to AFTER NAT in Prerouting and BEFORE NAT in Postrouting.
I do not know if my setup is common , but i have some like this:


DOWNLOAD LINK  [1024K]            -> HTB
          PEOPLE [500k-1024ceil]
                  guy1 [100k-1000ceil]
                  guy2 [100k-1000ceil]
                  guy3 [100k-1000ceil]
                  guy4 [100k-1000ceil]
                  guy5 [100k-1000kceil]

          ROUTER[512k-1000ceil]  -> router and P2P BOX 24/7  [Before be able
to shape the router , i need to hard limit the download speed to not eat all
bandwidth, now i can let it borrow if there is available bandwidth in the
parent (DOWNLOAD LINK)] -> I really need this 512k rate guaranteed in the
router.

          DEFAULT[8k-8k]  -> In theory should not be used by anyone :)



But now i have a doubt , when a packet gets in too htb it will be queued ,
right ? If it is, is there a way to drop it if it is over the ceil limit ? I
really do not want packets being queued `cause probably wiill  delay the
interactive traffic. Despite that , my setup is working great , that is just
a doubt :)


-- 
[]'s
Salatiel

"O maior prazer do inteligente é bancar o  idiota
   diante de um  idiota que banca o inteligente".
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.ds9a.nl/pipermail/lartc/attachments/20070423/4172234b/attachment.htm