[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[altq 112] Re: YOD (yet other doubts...)
Lin Su Mei wrote:
>> I know that. The program should implement like what you said but actually it
>> is not. That is, the
>> program doesn't matches what it said on paper and it doesn't matches the
>> spirits of CBQ
>> =>"in conjection time(when a class becomes regulated) a class will get it's
>> allocated link-sharing
>> bandwidth withouth punishedment from how much it used before. "
The original design of CBQ allows to control the penalty for excess
bandwidth use in the past.
You can set "minidle" to 0 not to punish a class.
From A.1 of the CBQ TON paper:
The implementation of the estimator should also explicitly
consider to what extent the limit status of a class should be
influenced by previous bandwidth the class has received in
excess of its allocated share. In our simulator, this is done
with a parameter that specifies a minimum (negative) value for
the variable avg.