[altq 64] Re: ATM Drivers with CBQ problems resolved?

[Kenjiro Cho <kjc@csl.sony.co.jp>]

Stewart Fallis <stewartf@mirror.bt.co.uk> said:
>> In the previous version of AltQ (was possibly a driver problem) there were some 
>> issues about using ATM interface cards as these cards provided some on-card 
>> buffering that interfeared with CBQ's attempt to effectively schedule incomming 
>> packets.  

>> Has this been resolved as I see on the web page there are updated ATM drivers,
>> do these drivers address this problem?

Cell scheduling in ATM basically interferes with packet scheduling in
ALTQ.  For example, if UBR or VBR is used, it is hard to predict the
interference between ATM and ALTQ.  On the other hand, ALTQ over CBR is
straightforward, and ALTQ assumes only this scenario.

The ATM driver update in altq-1.2 is to support multiple CBR PVCs on
a single physical device.
Before altq-1.2, pvc shadow interfaces on the same physical device
share a single output queue structure assigned to the physical device.
Thus, it was not possible to enable ALTQ on multiple pvc shadow
interfaces on the same device.  We had to use one ATM card for each
PVC.

In altq-1.2, each pvc sub-interface (renamed from a shadow interface)
has an independent output queue structure, and ALTQ can be enabled on
multiple pvc sub-interfaces on the same device.
Now, we can configure multiple CBR PVCs on a single ATM card, and
enable ALTQ on those PVCs.

-Kenjiro