Why does my computer experience high memory usage while attempting to run "parsim" multiple times?

Why does my computer experience high memory usage while attempting to run "parsim" multiple times? My "SimulationOutput" data is replaced every run to manage memory but I still experience high memory usage.

Answer 1

[MathWorks Support Team]

If the program involves running "parsim" several times (in a loop or otherwise), the client in the parallel pool slowly collects data from the workers and then converts it into a "SimulationOutput" object at the end of each simulation run. A good practice to follow here would be to remove the output data of a previous run from memory. However, even then, at a certain point in time, it is possible that the "SimulationOutput" from at least two separate runs exist simultaneously in memory, i.e., the current run's outputs in the function workspace and the previous run's outputs in the base workspace. It is important, hence, to monitor the size of the "SimulationOutput" one is dealing with.

Here are a few workarounds to consider regarding the above issue:

1. Using fewer workers is one option. It is recommended to have at least 8 GB of memory per worker for Simulink workflows. Please refer to the following link for more information:https://www.mathworks.com/support/requirements/matlab-parallel-server.html
2. Limiting the size of the logged data will help. For more details on this topic, execute the following command in the MATLAB R2019b command window:

   >> web(fullfile(docroot, 'simulink/ug/limit-amount-of-exported-data.html'))

3. Using a "postSimFcn" to delete redundant data from the output is another way to control this issue. To learn how to set a "postSimFcn", run the following command in the MATLAB R2019b command window:

   >> web(fullfile(docroot, 'simulink/slref/setpostsimfcn.html'))

4. Use the "Log Dataset to file" option as described in the documentation. For release-specific details, execute the following command in the MATLAB R2019b command window:

   >> web(fullfile(docroot, 'simulink/gui/log-dataset-data-to-file.html'))

It might seem like the below line of code would also achieve the same goal as the "Log Dataset to file" option:

in = in.setModelParameter('LoggingToFile', 'On')

However, this will not have much of an effect if the Simulink model uses "To Workspace" blocks. These blocks do not support the "Dataset" format, which is the format in which data is stored by the "Log Dataset to file" option. Instead of the "To Workspace" block, the workaround here is to send the output to "outport" blocks along with using the "Log Dataset to file" option as linked above.

Please follow the link below to search for the required information regarding the current release:

https://www.mathworks.com/help/