Gurobi freezes after presolving

[b Yang4869]

Hi Gurobi community,

I'm solving a LP. However after presolving, Gurobi freezes. I've waited for more than 1 hour, but nothing happens. Here is the log:

Optimize a model with 160000 rows, 94364 columns and 591380593 nonzeros

Coefficient statistics:

  Matrix range     [1e-06, 7e+04]

  Objective range  [1e-05, 1e-05]

  Bounds range     [0e+00, 0e+00]

  RHS range        [7e-05, 2e+03]

Presolve removed 0 rows and 0 columns (presolve time = 24s) ...

Presolve removed 0 rows and 5517 columns (presolve time = 25s) ...

Presolve removed 0 rows and 5517 columns (presolve time = 49s) ...

Presolve removed 0 rows and 5517 columns (presolve time = 71s) ...

Presolve removed 0 rows and 5517 columns (presolve time = 225s) ...

Presolve removed 0 rows and 5517 columns (presolve time = 235s) ...

Presolve removed 0 rows and 5517 columns

Presolve time: 235.23s

Presolved: 160000 rows, 88847 columns, 591380593 nonzeros

Could you tell what's going on with Gurobi and how could I avoid this issue?

Best,

Bo

[Bharath ae15d019]

Maybe there are too many variables.

--

---
You received this message because you are subscribed to the Google Groups "Gurobi Optimization" group.
To unsubscribe from this group and stop receiving emails from it, send an email to gurobi+un...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[b Yang4869]

Hi Bharath,

Thank you for your response, but the model is actually not very big. I'm running Gurobi on a server so I think both the machine and Gurobi should be able to handle problem with such size.

Best,

Bo

在 2018年8月12日星期日 UTC-4上午8:38:46，Bharath写道：

[b Yang4869]

I'm using C++ and Gurobi 7.5

在 2018年8月12日星期日 UTC-4上午8:36:01，b Yang4869写道：

[Daniel Espinoza]

Hi,

Note that just in non-zeros for the original problem, you have 8G of ram, add another 8G for the presolved model, and then you need some memory for the algorithms.

I would try on a machine with at-least 32G of ram... and watch the memory usage closely. Remember that as soon as you start to use SWAP, the performance goes over the board.

Using just one thread will reduce memory consumption for sure (Threads=1).

Just out of curiosity..... in average you have around 4000 nz for each constraint... which is rather unusual; what are you modeling?

Best,

Daniel

[b Yang4869]

Hi Daniel,

Thank you for your response. The machine has 128G RAM which is sufficient for this problem. I'm using a linear combination of basis functions to approximate some value, so the number of variables in each constraint is large. 

So basically do you mean Gurobi freezes because the model is too large?

Best,

Bo 

在 2018年8月13日星期一 UTC-4上午7:34:29，Daniel Espinoza写道：

[Daniel Espinoza]

Hi Yang,

No, I am not saying that... still, it's fairly large, and for sure, I would try to run with threads=1 or method=2 and watch the memory usage while its going, barrier tend to be more memory intensive than simplex.

Also, the matrix coefficients seem to have a wide range... that may cause issues too, and the objective function coefficients seem too small, I would bump them up by two orders of magnitude.

Could you share this model?