6
$\begingroup$

I'm solving a large instance using Gurobi 7.0, 64-bit version on machine with 16GB of RAM. Gurobi is stuck at the start and all the memory is exhausted. Here is the output:

5378534 variables, all binary
16140318 constraints, all linear; 319218593 nonzeros
        77 equality constraints
        16140241 inequality constraints
1 linear objective; 4842072 nonzeros.

#presolve       75.2705         1006633000      20529205440
#output         200.524         268435464       20797640904
Gurobi 7.0.0: timelim=3600
threads=2
nodefilestart=0.5
outlev 1
Optimize a model with 16140318 rows, 5378534 columns and 319218593 nonzeros
Variable types: 0 continuous, 5378534 integer (5378534 binary)
Coefficient statistics:
  Matrix range     [1e+00, 1e+03]
  Objective range  [1e+00, 4e+03]
  Bounds range     [1e+00, 1e+00]
  RHS range        [1e+00, 7e+04]
Found heuristic solution: objective -0

I was expecting out of memory error, but Gurobi is just stuck for hours, and CPU usage is 0.

Is this normal behavior for this instance size? Why am I not getting any errors?

$\endgroup$
  • $\begingroup$ And what is the question? $\endgroup$ – dhasson Jan 19 at 17:41
  • $\begingroup$ Thank you for the comment, I added the questions at the end. $\endgroup$ – lazar Jan 19 at 17:47
  • 1
    $\begingroup$ Thanks for the information. I guess the question could be suited for support.gurobi.com/hc/en-us $\endgroup$ – dhasson Jan 19 at 20:56
  • 2
    $\begingroup$ this is a huge model, it "just" takes some time to solve the root node LP... (my guess) $\endgroup$ – Marco Lübbecke Jan 20 at 8:30
4
$\begingroup$

It might very well be the case that your machine is starting to "swap", i.e. the hard drive is used for additional memory since the RAM is already exhausted. This will also dramatically increase IO time so your CPU is barely working because it takes too long to read/write new data.

See here for more info: https://en.wikipedia.org/wiki/Paging

For such a huge model, you should definitely look for a machine with a lot more memory. And even then you should be prepared to wait a while until you see some progress.

I recommend you try to find a different way to model your problem - maybe there is a way to decompose it into smaller problems.

| improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.