Is this a good process for changing seed number in manual model runs

Question

question

Paul G8 asked Jun 9, '22 Phil BoBo edited Jun 9, '22

Is this a good process for changing seed number in manual model runs

I'm aware of how to set up the model to be controlled externally, and how to deal with the experimenter, including which seed it uses normally in manual runs ("Repeat Streams of Replication"). What Im looking at is a merge of the two if the experimenter wasnt useable.

Since there isnt a scriptable way of accessing the manual run seed number setting in Experimenter (the dropper at least doesnt show any settings when using it) - I wanted to see if there was a way of doing it with "randInit()" and "getStream()" at say, onModelOpen, and then letting the rest of the model do its thing. Eg.

int streamEg = getStream(current);
randInit(streamEg, <seedNumVariable>);

getStream(current) from the onMoelOpen errors out.

On a related point - would any call to random stream even work in a model being controlled externally - assuming it was being run via a batch file on an AWS instance where it wouldnt necessarily have a licensed version of the program?

Software Version:

FlexSim 22.1.2

experimenter external control

Answer 1 · 2022-06-09T14:14:44Z

Phil BoBo answered Jun 9, '22 Phil BoBo edited Jun 9, '22

If you want to programmatically set the replication number that the streams will use in a regular non-experimenter run, you can set a node in the Tools folder before resetting and running your model:

set(node("MAIN:/project/environment/settings/randomrepeat"), 1);
treenode repnrnode = assertsubnode(node("MODEL:/Tools"), "randomrepeatrep");
repnrnode.value = 2;

This is the value that the Experimenter UI sets.

This will only work with a valid FlexSim license. Without a license, the random streams are locked to a particular default seed. Each FlexSim instance will need to be licensed when using the method above.

When using the Experimenter's Distributed CPUs feature, only the FlexSim instance running the experimenter needs to be licensed. The cloud nodes running replications for that experiment will be licensed via the experimenter.

Answer 2 · 2022-06-09T09:11:12Z

Jason Lightfoot answered Jun 9, '22 Paul G8 commented Jun 9, '22

Use a licensed version of FlexSim to give jobs to the AWS instance via the webserver interface.

· 3

Paul G8 commented · Jun 09, 2022 at 09:30 AM

Sorry - for some reason I had it in my head that the Flexsim version installed on the webserver didnt require a license. To clarify is that laso the case if Im just connecting to other computers on the local network via Experimenter too?

0 ·

Jason Lightfoot ♦ Paul G8 commented · Jun 09, 2022 at 11:10 AM

The requirements for the remote computers are listed here: Running Jobs in the Cloud

0 ·

Paul G8 Jason Lightfoot ♦ commented · Jun 09, 2022 at 11:29 AM

Thanks. So running cases via the experimenter only needs a runtime license (that includes machines accessible via AWS). Running a model directly via the webserver or from the commandline will need a license.

So, assuming licensed installs of FlexSim, will the above code actually work for altering the seed setup in much the same fashion as the experimenter would do it?

0 ·

question