Path diagnosis

Independently measure all model parameters:

[MSMO, July'97 CCR]
The model should agree with the actual performance
Otherwise it suggests a non-network bottleneck or a TCP or instrumentation bug

Do all parameters agree with expected values?
Excessive RTT suggests incorrect routing or congestion
Excessive loss suggests flaky gear or congestion
Wrong MSS suggests a problem with MTU discovery

Direct instrumentation for some "pathological" events
Duplicated packets, burst losses, excessive reordering, etc