[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

CBQ bug??



In the CBQ.cc there is written:
1) deque: this gets invoked by way of our downstream (i.e. linkdelay) neighbor doing a 'resume' on us via our handler (by Queue::resume()), or by our upstream neighbor when it gives us a packet when we were idle,
2) When a class comes suspended comes mailed the field undertime (undertime_ = fin_time + tx_time * (1.0 / allotment_ - 1.0) - (1-POWEROFTWO) * avgidle).
But from the point 1) the class sends a new packet only when a new packet arrive in the queue (and now > undertime, if no borrowing). This involves an approximation of the value undertime, therefore the class instead of transmitting the own packets immediately after undertime, delays this operation. The result is that the class goes slower, the own theoretical limits are under esteemed.