Actually, I am realizing that this doesn't work quite the way it seemed to, it's a bit more complicated. Take my example above: upperlimit m1=100%, d=5min, m2=20%… well, the 5 min seems to depend on how much of the 100% the user is actually getting, it will be longer than 5 min if the user is only able to use say, 50% of the bandwidth because he is sharing with another user, not really understanding exactly how the timer works in fact. Maybe someone will explain it but seems like very few people really understand hfsc queues. Anyway, in this scenario, the users may never get throttled when there are lots of users "fighting" over the available bandwidth, instead they get throttled when no one is using it, which is the opposite of what I would hope for. Looks like back to studying and experimenting for me, rather than posting in the forum!