The example here (LOOP12.EST) is a loopback version of the SLAP protocol.
By using loopback, we greatly reduce the state space explosion problem.
The PEW can build the behaviour graph for this specification in about
five minutes, and can make accurate performance predictions for different
parametric configurations in about five minutes per configuration.

The SLAP transitions are t6 to t23. We first build the BG with:

EC LOOP12
EI LOOP12 -C 5

This takes a few minutes.

We then edit the file PARAM.DAT for each experiment. For example,
if we have a delay of four when the provider relays an I frame (transition
t5), and a timeout delay of 20 (transition t15), and the probability of
losing an I frame is 10% (transition t3), then we specify the parameters:

D5	4
D15	20
L3	.1

To obtain the prediction, we then note the maximum depth reached in
building the graph (which is printed in the file TREE). In this case,
it is 20. We thus use:

GP 20 6 23

to get the predicted transition throughputs for the SLAP process
(transitions t6 through t23),

GP 20 1 5

to get the predicted throughputs for the provider process (transitions
t1 through t5), and

GP 20 24 26

to get the predicted throughputs for the user process (transitions
t24, t25 and t26).

Note that the process can be speeded up considerably by using a
comparison depth of less than 20. 5 may even be good enough. 20
is the value that is guaranteed to be sufficient for that specific
graph, however. If it doesn't work with the greatest depth reached
in *building* the graph, then the graph probably needs to be built
with a larger node comparison depth as the depth used was inadequate.




