The Solver should show a feasible solution within a few seconds, at least a few tens of seconds in a nominal situation. The solving speed is essential for interactive use. This solving speed enables the lambda function on limited solving time in AWS.
Detail Result
The data shows the time to solve the optimal solution with proof. (Note; A meta-heuristic solver can’t prove it except for zero of the objective function value.)
Benchmark Instance | ScheduleNurseⅢ | Gurobi9.0.1 | CPLEX12.9.0.0 | StateOfArtRosteringSolver | Optimum Value |
---|---|---|---|---|---|
Ikegami-2Shift-DATA1 | 8sec | 5.18sec | 15.3sec | 202sec | 0 |
Ikegami-3Shift-DATA1.1 | 22sec | 375sec | 20405sec(1134sec found3) | At 1hour (The objective function value=13) | 3 |
Ikegami-3Shift-DATA1 | 15sec | 296sec | 3418sec(2293sec found 2) | At 1hour(The objective function value=13) | 2 |
Ikegami-3Shift-DATA1.2 | 16.7sec | 737sec | 15424sec(2014sec found3) | At 1hour (The objective function value=14) | 3 |
Professor Ikegami’s benchmark data sets are trendy in OR academies, basing on real-world scenarios. MPS files and project files are available on Github. See the site below for further references.