[LARTC] Shaping outgoing traffic over multiple devices

[LARTC] Shaping outgoing traffic over multiple devices

[Sebastian 'spax' Pape]

hi!

I searched the archives and found this question a few times, but I
didn't find any answer :o Also I didn't find any hint at the howto.

I want to shape outgoing traffic over multiple devices (let's say eth0
and eth1). If I create a root discipline I have to name a device.
Therefore it seems I need two root disciplines and borrowing won't be
able. So if I don't want to absolutely limit the devices (half by half
or so) eth0 and eth1 will take as many bandwidth as they can get
without caring of the other interface. Is there any solution like
setting up a tunnel or "abusing the teql device" or something else?

best regards
		Sebastian

-- 
Sebastian 'spax' Pape          | Never argue with an idiot. They drag you down
mailto: sebastian@p-a-p-e.de   | to their level, then beat you with experience.
gpg: http://p-a-p-e.de/gpg.asc |
         --- Do you want to know more? http://www.p-a-p-e.de/ ---

_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

Re: [LARTC] Shaping outgoing traffic over multiple devices

[Stef Coene]

On Tuesday 15 October 2002 22:05, Sebastian 'spax' Pape wrote:
> hi!
>
> I searched the archives and found this question a few times, but I
> didn't find any answer :o Also I didn't find any hint at the howto.
>
> I want to shape outgoing traffic over multiple devices (let's say eth0
> and eth1). If I create a root discipline I have to name a device.
> Therefore it seems I need two root disciplines and borrowing won't be
> able. So if I don't want to absolutely limit the devices (half by half
> or so) eth0 and eth1 will take as many bandwidth as they can get
> without caring of the other interface. Is there any solution like
> setting up a tunnel or "abusing the teql device" or something else?
Yes, the imq device.  This is a virtual device (you can have more then one) 
and you can add a qdisc to it.  You can redirect packets to it with iptables 
and this can be done on each interface and for in and outgoin packets.
In your case you have to create 1 imq device and redirect all outgoin packets 
to it and shape the imq device.
More info http://trash.net/~kaber/imq/

Stef

-- 

stef.coene@docum.org
 "Using Linux as bandwidth manager"
     http://www.docum.org/
     #lartc @ irc.oftc.net

_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

Re: [LARTC] Shaping outgoing traffic over multiple devices

[raptor]

|Yes, the imq device.  This is a virtual device (you can have more then one) 
|and you can add a qdisc to it.  You can redirect packets to it with iptables 
|and this can be done on each interface and for in and outgoin packets.
|In your case you have to create 1 imq device and redirect all outgoin packets 
|to it and shape the imq device.
|More info http://trash.net/~kaber/imq/
]- does later packets go to the device they was directed in the begining.... i.e. if packet has to go out trought eth1 but i redirect it to imq, the flow is shaped to the desired bandwith but after that it continues its normal path.... i.e. flow become from this :

eth0  <----> eht1

this :

eth0 <--> imq <--> eth1
_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

Re: [LARTC] Shaping outgoing traffic over multiple devices

[Stef Coene]

On Wednesday 16 October 2002 15:11, raptor wrote:
> |Yes, the imq device.  This is a virtual device (you can have more then
> | one) and you can add a qdisc to it.  You can redirect packets to it with
> | iptables and this can be done on each interface and for in and outgoin
> | packets. In your case you have to create 1 imq device and redirect all
> | outgoin packets to it and shape the imq device.
> |More info http://trash.net/~kaber/imq/
>
> ]- does later packets go to the device they was directed in the
> begining.... i.e. if packet has to go out trought eth1 but i redirect it to
> imq, the flow is shaped to the desired bandwith but after that it continues
> its normal path.... i.e. flow become from this :
>
> eth0  <----> eht1
>
> this :
>
> eth0 <--> imq <--> eth1
I's more like
eth0 -> imq0 -> forwarding -> eth1
eth1 -> imq0 -> forwarding -> eth0

You can even do
eth0 -> imq0 -> forwarding -> imq0 -> eth1

One of the disadvantage is the delays.  The packet will travel thru 2 queues.  
So shaping on imq AND ethx is not such a good idea.

Stef

-- 

stef.coene@docum.org
 "Using Linux as bandwidth manager"
     http://www.docum.org/
     #lartc @ irc.oftc.net

_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/