Multiple instances of the same Gurobi model executing in parallel

[dccoo...@gmail.com]

One of my usecases would have a large number of optimisations occurring in parallel with some tight wallclock requirements.

There's a useful reference to parallelising a Gurobimodel in a master-worker model:

http://www.gurobi.com/pdfs/webinar-parallel-and-distributed-optimization-english.pdf

and the reference manual has a section on Distributed Parallel Algorithms.

The parallelisation described there is either parallelising the computation for a single model or trying multiple strategies in parallel on the one model.

I would like to have multiple instances of the same model executing in parallel with different values of some constraints and decision variable coefficients.

Has anyone tried this as either one optimisation model per process (job parallelisation) or parallelising model instances within a process using, say, openmp?

Thanks

David

[Tobias Achterberg]

In such a case you should implement the parallelism yourself with whatever tools your
programming language offers. Just be sure that for each thread you need to create an
individual Gurobi model. You cannot work on the same model object concurrently.

Regards,

Tobias