Increase number of parallel scenarios run with behaviour space

44 views
Skip to first unread message

Barak Garty

unread,
Apr 25, 2024, 1:34:27 PMApr 25
to netlogo-users

I am using netlogo 6.4.0. When I set a runs through behaviourSpace it reads that it can run in parallel between 18 and 24 scenarios. I read that I can change the ram amount it has access to in the netlogo cfg files and restart for better performance of the interface, presumably increase number of parallel runs. I changed the netlogo cft file from 50% to 80%. There is no guide that indicates which of the 4-5 cfg fules should be changed. I was hoping or thinking that changing that file the way I did meant that now it has access to 80% of my pc ram (16GB). However, after saving that change and restart netlogo, it still saying I can only run 18-24 scenarios in parallel.

Sorry for the long explanation. Does anyone know if I am doing it wrong? Is this method if allocating more ram is the way to increase the number of parallel runs?

Thanks.


Aaron Andre Brandes

unread,
Apr 29, 2024, 4:51:46 PMApr 29
to Barak Garty, netlogo-users

Hi

The value for recommended maximum number of maximum runs is a heuristic guideline which we developed testing on a system with the 50% RAM figure. In our test the amount of memory used per run was not a limitation, and the guideline was based on issues with garbage collection and interactions between the cores.

You can experiment with using more parallel runs. If you are getting better overall throughput then what you are doing is fine. It’s also possible, depending on your experiment that you don’t need all that extra memory.

 

You can use a tool like VisualVM if you want to explore NetLogo’s memory use when running BehaviorSpace

(see https://www.baeldung.com/java-profilers for more information about Java Profilers)

 

I think Michael Tamillow’s suggestion of running heedlessly is a good idea.

 

Let us know if you have any further questions.

  • Aaron

 

-- 

Aaron Brandes, Software Developer

Center for Connected Learning and Computer-Based Modeling

 

 

--
You received this message because you are subscribed to the Google Groups "netlogo-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to netlogo-user...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/netlogo-users/6f9a0a15-d09e-4b79-8098-e9decc58a4c4n%40googlegroups.com.

Aaron Andre Brandes

unread,
Apr 29, 2024, 5:18:41 PMApr 29
to Barak Garty, netlogo-users

Hi,

I think my answer was not accurate. The recommended maximum is actually the total number of processors for your system.

In your case that is 24, which may correspond to 12 dual core CPUs. It is possible to request more parallel runs than cores. We did not test performance with more than the total number of processors.

The default value is what we chose as a guideline based on some testing. It is determined by rounding down ¾ of the number of processors to the nearest integer.

My advice to experiment and to try headless runs is still appropriate.

 

An easy way to find the total time taken by the experiments is to add the following:

 

Pre experiment commands:

reset-timer

 

Post experiment commands:

show timer

 

The results will show up in the command center if you are running with the GUI, otherwise you need to write to a file.

 

If you want the overall experiment to be reproducible (the better to compare performance) you can use the following to set a different random seed for each run.

 

Setup commands:

random-seed (474 + behaviorspace-run-number)

setup

 

- Aaron

Barak Garty

unread,
May 4, 2024, 4:46:01 PMMay 4
to netlogo-users
Hi  Aaron,

 Much apologies for a very late reply. Was of-greed for a while. 
I hired VM from Azure to help me run my model as my home PC is not good enough. I rented a win11 OS 64cpu 128GM VM.
It shows me now that it can run 48 to 64 runs in parallel, when I set an experiment from the behaviourSpace UI. 
I like to ask you maybe you know, which CFG file I need to alter this line (
java-options=-Xmx####m) , to get more RAM usage out of the RAM available on the VM?
I can see the following CFG files:
Behaviorsearch.cfg
HubNet Client.cfg
NetLogo 3D.cfg
NetLogo.cfg
NetLogo_Console.cfg

In the Netlogo guide there an explanation on how to set an increase RAM, but not which CFG file belongs to the behavioureSpace if I run it through the NetLogo GUI.

Kind regards,
Barak

Aaron Andre Brandes

unread,
May 6, 2024, 1:48:59 PMMay 6
to Barak Garty, netlogo-users

Hi Barak,

If you are running a regular BehaviorSpace Experiment  in a Linux VM (or on a MacOS or Linux machine) you want to change the NetLogo.cfg. For a Windows VM you would need to change the NetLogo_Console.cfg file. You would change the NetLogo 3D.cfg file for a NetLogo 3D BehaviorSpace experiment on Linux or MacOS.

Reply all
Reply to author
Forward
0 new messages