Dare wrote: > Hello, > > I need to limit the bandwidth to 1Mbit per user IP (and never give > more) and inside of each 1Mbit prioritize minimum-delay traffic (video > stream from specific servers, so I mangle TOS for packets from those). > The problem is I need this for *thousands* of users. I am guessing > thousands of classes are needed, but the question is, is this > feasible? Could one machine ever handle this? How does this scale with > either CBQ or HTB? > > thanks, > Darko > > ps. in more detail: I am trying to do this before (upstream) an ATM > switch, which has the bw limit per user set. If I just prioritize the > traffic before reaching the switch, other traffic still gets through > and the switch is the one that drops the packets on its own, not > knowing about my priorities. That's why I think I need to both shape > and prioritize the traffic in the same point and while still on the IP > level. > > _______________________________________________ > LARTC mailing list / LARTC@mailman.ds9a.nl > http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/ > hi, if your switch has the bw limit per user set, perhaps you only need to set one prio queue for minimum delay traffic. if you prefer to shape to 1Mbit and have up to 1000 classes htb is certainly better to do this versus CBQ ; but feasibility will certainly depend on number of backlogged classes, the way you filter packets, kind of driver you use (interrupt mitigation ) average number of paquet/s and of course CPU frequency. -- Vincent EGAL Email : egal@ipanematech.com _______________________________________________ LARTC mailing list / LARTC@mailman.ds9a.nl http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/
