Matlab parpool gives me 4 workers on Azure VM instead of 8 workers

I just deployed Matlab 2024b on Azure VM, with 8gb vCPU, 32gb RAM. However when i start the parpool with

% Start a parallel pool
parpool('local' );
parfor linearIdx = 1:totalCombinations
...
end

it runs with 4 workers whereas i would expect 8 workers since i have 8 vCPUS.

also parcluster('local') gives me

 Local Cluster

    Properties: 

                   Profile: Processes
                  Modified: false
                      Host: matlabvm
                NumWorkers: 4
                NumThreads: 1

        JobStorageLocation: /home/ubuntu/.matlab/local_cluster_jobs/R2024b
   RequiresOnlineLicensing: false
   PreferredPoolNumWorkers: Inf

    Associated Jobs: 

            Number Pending: 0
             Number Queued: 0
                Number Running: 1
           Number Finished: 0

Please note my Matlab 2024b licence is a Trial version. Could this be the case? Thanks

Answer 1

[Pratyush Swain]

Hi Farouk,

By default Parallel Computing Toolbox will attempt to use a number of workers equal to the number physical cores of the computer. In a cloud environment, a vCPU typically represents a portion of a physical CPU core.

To programmatically alter the maximum number of workers allowed by the 'Processes' or 'local' profile, use:

c = parcluster('Processes')
c.NumWorkers = 8;
saveProfile(c) % will overwrite the profile with the new NumWorkers value

Please also refer to http://www.mathworks.com/matlabcentral/answers/2042101-how-can-i-increase-the-number-of-matlab-workers-inside-parallel-computing-toolbox-to-use-all-the-log for information on why the number of cores detected might not be all of the cores on your machine.

Hope this helps.

[Farouk]

Hi Pratyush,

Thanks yes you are right. i later figured it out and did it exactly this way and it worked. Cheers.