[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Q on RED: is RED broken?
Hello..I will appreciate if anybody answers this question.
I don't understand why RED produces the following behavior.
I use 3 flows agrregated into one RED queue.
Each flow generates UDP traffic packets with the same length that
I designed as follows.
At each time step, no packet is generated, or 3 packets are generated.
(One flow can generate at most 1 packet per step.)
At each time step, the arrival times of all 3 packets are exactly identical.
(My traffic trace file is the evidence of this.)
The expected behavior of RED must be like this.
It calls enque function 3 times and then calls deque. (enque -> enque ->
enque -> deque.)
But sometimes it calls enque -> deque -> enque -> enque -> deque.
This happends quite often.
This behavior is critical to simulate, e.g., a buffer-size 1 policy.
(min th = 1, max th=1, drop probability = 1.)
Does this mean RED is broken ?
Hyeong Soo Chang.