Using Windows/Linux

進行中

Theresa Reinhardt

• 2021年06月21日 16:09

Hello.

Gurobi needs about 5 hours to solve a MIP model on a 6-core-machine on Windows. On a cluster (6 cores, Linux) the same model takes about 33 hours until its done. Furthermore, an easier MIP-model is done within 10 minutes on both operation systems. Is there any comprehensible reason for these significantly different calculation times in the first case?

Thank you.

 

0

• 正式なコメント

  

  Simranjit Kaur

  Gurobi Staff

  • 2024年12月06日 00:30

  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?.
• 

  Sonja Mars

  Gurobi Staff

  • 2021年06月23日 06:22
  • 編集済み

  Hi Theresa,

  Can you share some log files or snippets from the log files (I guess the 33 hours log might be a bit long for a post)? This would be helpful to understand a bit better what is going on.

  In general, the solver can take a different path through the branch-and-bound tree on different operating systems. And a different path can result in a different runtime. 

  Best,
    Sonja

  0
• 

  Theresa Reinhardt

  • 2021年06月22日 13:43

  

  

  

  Hi Sonja.

  Actually, I can not directly interact with gurobi as there is an other framework in between. So to write the logfile I passed the parameter 'LogFile' (the same way I passed the parameters 'NodefileStart' and 'MIPGap' which work). But it is not writing a logfile.

  You say the runtime can depend on the operating system. So it seems like Gurobi might be more suitable for Windows in general or just in these specific case?

   

  0
• 

  Sonja Mars

  Gurobi Staff

  • 2021年06月23日 06:21

  What tool are you using? Did you provide a full path to the log file location? It is possible that you have to additionally set something like outputflag. 

  You say the runtime can depend on the operating system. So it seems like Gurobi might be more suitable for Windows in general or just in these specific case?

  No, the solver can just take a different path. If you use a different data set for your model it can be the opposite. You might find this article useful: Why does Gurobi perform differently on different machines?

  Without a log file it is really hard to say what exactly is going on. 

  Best,
    Sonja

  0
• 

  Theresa Reinhardt

  • 2021年06月23日 13:00

  

  

  

  I am using the energy modelling framework oemof which uses the framework pyomo itsself. Providing the path and setting the output flag does not make it write the logfile. So there might be some issue with pyomo.

  Greetings

  Theresa

   

  0

投稿コメントは受け付けていません。