Terminated By Solver AIMMS

[Kasina]

Hi Guys,

I am solving a huge MIP (~ 50,000 Integer Variables). About my model, briefly, It is a two stage model with multiple scenarios - hence the large number of Integer variables. Each scenario has 100 Hours worth of data. The solver status says "Terminated by Solver" while the gap is still 16.42%. Why does this happen. I set the relative gap to 0.03% from CPLEX MIP settings. With such a huge model, the solver should take quite some time to solve. But just after 2000 secs (~30 mins), I get this message and the model stops running. It doesn't even explore any nodes. I checked if the iterations reached their limit, but this doesn't seem to be the case. What might have happened ?
I'll appreciate any ideas on this. Thanks ! I am attaching a screenshot.

Kasina

[Guido Diepen]

Hi Kasina,

please take a look at the messages.log of AIMMS. If the solver is terminated for some reason, the messages.log most of the times contains some additional information.

You can view the messages.log by clicking Menu > Window > Messages or by pressing the keyboard shortcut Ctrl-m.

The information from this log might give a pointer to what is the reason for the solver stopping.

Guido Diepen

AIMMS Specialist

[Kasina]

Hi Guido,

Thanks for the advice. As suggested by some previous posters who
encountered this problem, I checked the iteration limit, but that
wasn't the problem. And as you suggested, I checked the message log
window. It says,"Postsolve: some variables were outside their bounds.
Set option postsolve_continuous_variables equal to 'Round to nearest
bound, resolve LP' and rerun the model to get a feasible solution with
all variables inside their bounds.
Warning: After 210044 iterations CPLEX 12.4 found an integer solution
to MathProg_AS1.
The minimum found for OFAS1 is 1.53796441064e+020."

(I expect the correct solution to be of e^10 or e^11 order)

I did observe that while the gap persists, it is different for
different runs. The run is "terminated by solver" at different gaps.
For example: Yesterday as I posted, it stopped at a gap of 16.42%.
Today, the gap was a 100%. This leads me to believe that this indeed
is some sort of a memory allocation problem although the message log
doesn't say so.

Any ideas? Thanks.

Kasina

[Guido Diepen]

Hi Kasina,

my first idea would indeed also be a memory allocation issue. Can you please also instruct the project to write a solver listing file? You can do this by setting the  

Menu > Settings > Project Options ... > Solvers general > Standard reports > Solver specific > Solver listing messages 

option to the value All. After running your project again, there should be a status file for the solver you used (in your case it will probably be a file named CPLEX 12.4.sta in the log directory of your project. Can you verify if there is more information in this log file?

Guido Diepen

AIMMS Specialist

[Kasina]

Hi Guido,

You're right. The STA file had this:

"Compressing row and column files.
Warning: MIP starts not constructed because of out-of-memory status.
Memory in use by CPLEX 12.4.0.0: 342531720 bytes."

So, any ideas on how to deal with this memory allocation problem?
Thanks!

Kasina

[Guido Diepen]

Hi Kasina,

not sure if you are using a 32 bit or 64 bit version of AIMMS at the moment. If you are using a 32 bit version, you could try to use the 64 bit version of AIMMS. Using the 64 bit version of AIMMS will ensure you are not running into the memory limitations of the 32 bit platform (i.e. 2GB / 4GB max memory per process, depending on whether your OS is 32 or 64 bit).

You don't need an additional license to run the 64 bit version of AIMMS, you just can download it from our website and install it. It should then automatically pick up the license installed used by the 32 bit AIMMS.

Guido Diepen

AIMMS Specialist

[Mayank Patel]

Hi Guido,

 

I was following this thread as I have a similar issue,

 

I am running the 64 bit version on a 64 bit operating system. After interrupting the solution (at around ~50%), the CPLEX 12.4.sta file gave the following message:

 

Memory in use by CPLEX 12.4.0.1: 247452704 bytes.

Postsolve: some variables were outside their bounds. Set option postsolve_continuous_variables equal to 'Round to nearest bound, resolve LP'
and rerun the model to get a feasible solution with all variables inside their bounds.

 

After changing the suggested option, I re-ran the model to recieve the same message again.

 

Any ideas?

 

Thanks

 

Mayank

[Guido Diepen]

Hi Mayank,

the only thing I can think of is that your model is not very well scaled. If this is the case, the solver could run into issues with numerical stability. Can you verify if you have very large and/or very small coefficient values in your model? If you do, please try to rescale these large/small values.

Guido Diepen

AIMMS Specialist