Originally posted by: SystemAdmin

Hi Daniel,

I have out of memory issues when I run my code with big data set (100 jobs) in the environment (Matlab 64 bit 2012a with Cplex 12.5 64 bit).
It was fine when the data set is small. When I test the 100 job instants, the used memory was cumulatively increasing Even my desktop has 16G memory, it showed "out of memory" and then clash down.

But when I run the same code and data set in my laptop (2010a Matlab 32bit with Cplex 12.2 32bit), I don't have the same issue. The used memory is always constant, but because it is old computer, it just slow.

Do you have any idea how to solve this issue? Thanks.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Do you perform repeated solves or do you just setup the model and solve it once?
Could you provide log files for the solves with your old and the new computer? Did the problem solve to optimality on the old machine?
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Hi,

It happened to all the machine.
The error code is
'{Error using cplexlink125
Out of memory. Type HELP MEMORY for your options.

Error in C:\Program Files
(x86)\IBM\ILOG\CPLEX_Studio125\cplex\matlab\x86_win32\@Cplex\Cplex.p>Cplex.addRows
(line 2331)
Error in C:\Program Files
(x86)\IBM\ILOG\CPLEX_Studio125\cplex\matlab\x86_win32\@Cplex\Cplex.p>Cplex.subsref
(line 3846)
Error in FTWCplexcode (line 37)
cplex.addRows(1, machinecons, 1); }'
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

It may be that the problem is just too big for your machine. There are several things you can try to reduce the memory requirements:
1. Enable memory emphasis.
2. Reduce the number of threads (see parameter CPX_PARAM_THREADS), in an extreme case use only 1 thread.
3. When you run out of memory during the solve then use node files on disk.
In your case however it looks like the out of memory happens while the problem is setup and before the solve does even start. Is that correct? If yes then there is basically nothing you can do (you can always get a bigger machine). The matlab connector is known to have some memory leaks. They will be fixed in the next fix pack. Things may get better when this is out but I cannot make any announcements when this will be released.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Thank you very much for your help.

Yes, it ran out of memory when it is building the model and before the solve started. I will try to ran it in a bigger machine instead.
When you mean bigger, do you consider the RAM or Processor?

The matlab connector is known to have some memory leaks. So can I avoid this memory problem if I connect CPLEX with Java without using bigger machine?

Thanks.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

> Yes, it ran out of memory when it is building the model and before the solve started. I will try to ran it in a bigger machine instead.
> When you mean bigger, do you consider the RAM or Processor?
>
I mean "more RAM".

> The matlab connector is known to have some memory leaks. So can I avoid this memory problem if I connect CPLEX with Java without using bigger machine?
>
Yes, this may help (although I cannot promise that the model built in Java will fit into memory).
The API with the smallest memory requirements is the C callable library. So if you want to save as many memory as possible and know some C then you can also try to code in C. However, if the problem barely fits into memory then trying to solve it will most probably run out of memory rather quickly.
#CPLEXOptimizers
#DecisionOptimization

Originally posted by: SystemAdmin

Thanks. Daniel. I do appreciate your help.
#CPLEXOptimizers
#DecisionOptimization