Simulink design optimization with Parallel computing toolbox

Hello all, I am using Simulink design optimization with parallel computing toolbox to tune some parameters( usually 10-15), the problem is with how many ever cores I start the optimization, it gives me a error saying

 'The session that parfor is using has shutdown' 
or 
 'The client has lost connection to Lab X' X being some number.

This occurs when I deal with pretty huge data sets.

from comments of Jason Ross to this answer http://www.mathworks.com/matlabcentral/answers/27749-cores-vs-speed-tradeoff-for-a-matlab-computer

the peak memory usage is 3.2 GB, my machine settings are Windows 7, Dual Quad(2.13GHz), 24 GB RAM, 240 GB SSD with 60GB page file and 70GB free space.

I hope I am clear enough, any help to resolve this issue would be appreciated

Answer 1

[Alec Stothert]

Hi Srinivas,

Is this for an estimation problem, in other words is it the expected model output data that is "pretty huge"? Can you send the data (or similarly sized data) to the MATLAB pool using a parfor loop, i.e., outside of SDO.

-Alec

[Srinivas]

Hi Alec,

This is an estimation problem, the expected model output data is "pretty huge", I have no problems running it on smaller data sets. I am not sure what you mean by sending the data to the MATLAB pool using a parfor loop.

[Alec Stothert]

Without knowing the whole estimation problem I was wondering if you hit a memory issue causing the pool to fail simply by using a parfor loop. So something like:

data = LoadMyHugeData

parfor ct=1:10

result(ct) = mean(data); %or some action on the data

end

I'll send you an email and we can discuss more off line.

-Alec