Solving the same MILP problem goes quickly on a laptop, but takes ages on the cluster, which should be faster?

Hi,

I'm having a problem that Gurobi (8.0.0 and 9.5.0) works well on my laptop (2+2 cores, Win64, nothing fancy) for solving some MILPs (there are many examples, not just one problem). On the cluster, which I had imagined should be faster, it works for some problems, but for some it seems it gets stuck, see below. Any idea what may be causing this? It looks like it is doing something for 5000 s that maybe it just shouldn't do. Is there some set of parameters stored somewhere that could be different on the local computer and cluster? The code I run is the same. I'm running this from matlab, gurobi v. 9.5.0. I'm here setting the mipgap param to 0.0030, in most cases I want it smaller.

Optimize a model with 7569 rows, 16634 columns and 41014 nonzeros
Variable types: 15408 continuous, 1226 integer (1226 binary)
Coefficient statistics:
  Matrix range     [1e-04, 2e+03]
  Objective range  [1e-02, 1e+03]
  Bounds range     [1e-04, 1e+03]
  RHS range        [0e+00, 0e+00]
Presolve removed 1815 rows and 7276 columns
Presolve time: 0.08s
Presolved: 5754 rows, 9358 columns, 29029 nonzeros
Variable types: 8249 continuous, 1109 integer (1102 binary)
Presolve removed 770 rows and 822 columns
Presolved: 4984 rows, 8536 columns, 27377 nonzeros


Root relaxation: objective -4.095203e+04, 4867 iterations, 0.12 seconds

    Nodes    |    Current Node    |     Objective Bounds      |     Work
 Expl Unexpl |  Obj  Depth IntInf | Incumbent    BestBd   Gap | It/Node Time

      0     0 -40952.027    0  102          - -40952.027      -     -    0s
H    0     0                    -40575.65622 -40952.027  0.93%     -    0s
      0     0 -40745.018    0   49 -40575.656 -40745.018  0.42%     -    0s
H    0     0                    -40582.46268 -40745.018  0.40%     -    0s
      0     0          -    0      -40582.463 -40745.018  0.40%     - 5000s
H    0     0                    -40582.77466 -40745.018  0.40%     - 5000s
H    0     0                    -40584.65622 -40745.018  0.40%     - 5000s
H    0     0                    -40587.45481 -40745.018  0.39%     - 5000s
H    0     0                    -40590.25341 -40745.018  0.38%     - 5000s
H    0     0                    -40591.26127 -40745.018  0.38%     - 5000s
H    0     0                    -40594.26127 -40745.018  0.37%     - 5000s
H    0     0                    -40596.26127 -40745.018  0.37%     - 5000s
H    0     0                    -40598.26127 -40745.018  0.36%     - 5000s
H    0     0                    -40600.26127 -40745.018  0.36%     - 5000s
H    0     0                    -40601.26127 -40745.018  0.35%     - 5000s
H    0     0                    -40603.26127 -40745.018  0.35%     - 5000s
H    0     0                    -40605.26127 -40745.018  0.34%     - 5000s
H    0     0                    -40607.26127 -40745.018  0.34%     - 5000s

Cutting planes:
  Implied bound: 50
  Flow cover: 34

Explored 1 nodes (4682627 simplex iterations) in 5057.82 seconds
Thread count was 32 (of 64 available processors)

Solution count 10: -40607.3 -40605.3 -40603.3 ... -40590.3

Time limit reached
Best objective -4.060726127389e+04, best bound -4.074501782888e+04, gap 0.3392%

Official comment

Simranjit Kaur

Gurobi Staff

November 16, 2025 23:43

This post is more than three years old. Some information may not be up to date. For current information, please check the Gurobi Documentation or Knowledge Base. If you need more help, please create a new post in the community forum, or try Gurobot, our chatbot interface offering instant, expert-level support.

Matthias Miltenberger

February 23, 2022 15:11

We will discuss this in an internal ticket.

Best regards,
Matthias

February 25, 2022 15:30

The issue was due to a problem with the cluster and not related to Gurobi.

Solving the same MILP problem goes quickly on a laptop, but takes ages on the cluster, which should be faster?

Comments

Didn't find what you were looking for?