I'm sorry for the long delay, for some personal reasons I'd been unable to reply
sooner :(

> Note that CBQ cannot really put hard limits on bandwidth.  Try HSFC for that.

I will try that today, but I am not sure if I really need to put a hard limit
on my bandwith.

> 
> What does altqstat show is happening with the traffic?  Do you see it
> being classified as expected?
> 

Yes, it does that. At least it looks okay to me.

Root Class for Interface pppoe0: root
        priority: 0 depth: 2 offtime: 1 [us] wrr_allot: 1562 bytes
        nsPerByte: 53333        (150.00Kbps),   Measured: 10.76K [bps]
        pkts: 52,       bytes: 27609
        overs: 0,       overactions: 0
        borrows: 0,     delays: 0
        drops: 0,       drop_bytes: 0
        QCount: 0,      (qmax: 30)
        AvgIdle: 2291 [us],     (maxidle: 2291 minidle: -77707 [us])
Default Class for Interface pppoe0: fast_class
        priority: 1 depth: 0 offtime: 67559 [us] wrr_allot: 2187 bytes
        nsPerByte: 76190        (105.00Kbps),   Measured: 9.32K [bps]
        pkts: 15,       bytes: 21930
        overs: 0,       overactions: 0
        borrows: 0,     delays: 0
        drops: 0,       drop_bytes: 0
        QCount: 15,     (qmax: 30)
        AvgIdle: 4464 [us],     (maxidle: 4464 minidle: -111011 [us])
Ctl Class for Interface pppoe0: ctrl_class
        priority: 1 depth: 0 offtime: 1422915 [us] wrr_allot: 312 bytes
        nsPerByte: 533333       (15.00Kbps),    Measured: 0 [bps]
        pkts: 0,        bytes: 0
        overs: 0,       overactions: 0
        borrows: 0,     delays: 0
        drops: 0,       drop_bytes: 0
        QCount: 0,      (qmax: 30)
        AvgIdle: 93749 [us],    (maxidle: 93749 minidle: -777082 [us])
Class 0 on Interface pppoe0: default_class
        priority: 1 depth: 1 offtime: 17361 [us] wrr_allot: 2812 bytes
        nsPerByte: 59259        (135.00Kbps),   Measured: 0 [bps]
        pkts: 0,        bytes: 0
        overs: 0,       overactions: 0
        borrows: 0,     delays: 0
        drops: 0,       drop_bytes: 0
        QCount: 0,      (qmax: 30)
        AvgIdle: 5231 [us],     (maxidle: 5555 minidle: -86342 [us])
Class 1 on Interface pppoe0: rest
        priority: 1 depth: 0 offtime: 632290 [us] wrr_allot: 625 bytes
        nsPerByte: 266666       (30.00Kbps),    Measured: 1.44K [bps]
        pkts: 37,       bytes: 5679
        overs: 0,       overactions: 0
        borrows: 0,     delays: 0
        drops: 5,       drop_bytes: 210
        QCount: 26,     (qmax: 30)
        AvgIdle: 41666 [us],    (maxidle: 41666 minidle: -388540 [us])


> With your slower outbound limit you might also try dropping the mssclamp
> value to about 500 bytes or so.  That should give more possible slots on
> the outbound wire for the ACK packets of your inbound connection to drop
> into.
> 

If I drop my mssclamp to 500 byte (in ipnat.conf), my download speed goes
down drastically. Even without altq enabled.

I will experiment a little bit more with the mssclamp value, maybe I
have to rewrite my ipnat.conf for it.

Bye
	Ralf