All,

I'm not sure if we can reach good test coverage if we apply more and more 
detailed "realistic" test cases, particularly if we are looking into the 
future. The more detailed the test cases are, the more relevant others we leave 
out.

Shouldn't we more concentrate on what we expect from a good AQM?

The only thing we might expect from an AQM is to prevent greedy TCP sources 
from drawing buffers permanently towards full state, instead of the much better 
almost empty state, while maintaining full link utilization. In all other 
situations, I'm convinced, AQM cannot improve a lot. But at least, in these 
cases, it should not make things worse.

Of course the greedy TCP case can be overlaid by others (unresponsive flows, 
application paced flows, a renewal process of short lived flows, synchronized 
start of flows etc.etc.), and we should take this into account. But we should 
not expect a lot if those "others" dominate the scenario.

Example: An overlay of finite flows arriving as a renewal process with inter 
arrival time T (seconds) and flow size S (bits). And a link capacity of C. In 
cases with C < T*S the link collapses with an infinite number of uncompleted 
flows, each of almost zero rate. In the opposite case, C > T*S the link cannot 
be loaded permanently at 100% since the traffic offer is simply not there. 
There is little to gain by AQM. It is only a thin range short before the 
collapse.

Wolfram

_______________________________________________
aqm mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/aqm

Reply via email to