Skip to main content

Exploring nodes is getting slower

Answered
Abdulkerim Benli
  • Gurobi-versary
  • First Comment
  • First Question

Hello,

As solving a large MIP instance, exploring the BB tree is getting slower in a certain depth. There is a screenshot below that clearly shows after some point even exploring a new node takes pretty much.  

The model has >1 million continuous and ~50 thousand binary variables, and it is run with default parameters. 

The running computer has 64 cores and 250 GB RAM where only 32 core is available with default Thread parameter.

I ve set NodeFileStart parameter as 25. 

I do not expect to solve it to the optimality, but getting a feasible solution in a reasonable time would be good.

Thank you in advance. 

 

0

Comments

5 comments

Date Votes
  • Official comment
    Simranjit Kaur
    • Gurobi Staff
    Edited
    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 why not try our AI Gurobot?.
  • Jaromił Najman
    • Gurobi Staff

    Hi Abdulkerim,

    If you are interested in a feasible point only, you should definitely try the No Relaxation Heuristic introduced with version 9.1.0, see NoRelHeurTime. You should set the time to a couple of hours to see whether a feasible point can be found this way. Other than that, you could also have a look at the list of Most important parameters.

    Best regards,
    Jaromił

    1
  • Abdulkerim Benli
    • Gurobi-versary
    • First Comment
    • First Question

    Thank you for quick response. NoRelHeurTime is working fantastic and it saved me. But, I just wanted to see how the BB will handle only with default Heuristics (0.05) and was curious about why node exploration is getting slower after some point. I should have been more clear, so sorry. Thank you again.

    0
  • Jaromił Najman
    • Gurobi Staff

    Node exploration may get slower because Gurobi tries to adjust and find a feasible point by using more expensive heuristics. Given the large objective value, it is also very likely possible that the problem gets more and more numerically unstable through branching resulting in an increased solution time when getting deeper in the B&B tree.

    0
  • Abdulkerim Benli
    Edited
    • Gurobi-versary
    • First Comment
    • First Question

    Thank you for the response.

    Now, it makes sense especially with the expensive heuristics part.

    I will also try to scale down the terms in the objective.

    0

Post is closed for comments.

Didn't find what you were looking for?

New post