Shaping is done on the physical device irrespective of number of virtual interface unless there is a specific virtual device support in kernel like for IMQ. Best option would be to shape traffic on eth1 and eth0. Activate policing for ingress queue on eth0 (external interface). Queue builds up for incoming on linux box for incoming traffic. Reason for this, no change/ recompilation of kernel required. Stable (look at recent post on IMQ bug fix). IMQ need be used only if you want to control/cap aggregated incoming and outgoing bandwidth and not control them individually. Mohan -----Original Message----- From: lartc-admin@xxxxxxxxxxxxxxx [mailto:lartc-admin@xxxxxxxxxxxxxxx]On Behalf Of Kai Weber Sent: Tuesday, June 24, 2003 6:20 PM To: lartc Subject: Re: [LARTC] traffic controlling strategy * alexandru matei <alex@xxxxx>: > 1/ using default outbound shaping on eth1 (outside interface) for > outgoing and outbound shaping on eth0 (inside interface) for incoming > 2/ using default outbound shaping on eth1 (outside interface) for > outgoing and inbound shaping (with IMQ) on eth1 (outside interface) for > incoming > ... > Which combination do you recommand? Any reason for your recommandation? I have the same problem and thinking of an third strategy: 3/ Splitting the outside interface into two virtual interfaces: eth0:0 for incoming, eth0:1 for outgoing traffic and routing the two interfaces that both traffic seems to go out of the interface. Bad english maybe, but probably totally wrong, too. Kai -- * mail kai.weber@xxxxxxxxxxx web http://www.glorybox.de pgp 0x594D4132 _______________________________________________ LARTC mailing list / LARTC@xxxxxxxxxxxxxxx http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
