Improve Oprimizer performance by using timeout settings, isolating inputs, and defining your decision variables.
While the Optimizer process is running and searching for the optimal solution, the Anaplan model is locked. To avoid impacting production models during model development, we recommend testing in a sandbox or test environment. If a problem is so complex that the Optimizer runs for a long time, consider separating the Optimization model.
When you set up Optimizer, you must specify a time-out setting in seconds to prevent the action from running indefinitely until canceled. The default setting is five minutes (300 seconds) and the maximum setting is six hours (21,600 seconds). An optimizer action always ends after six hours, even if you specify a longer time-out.
For performance analysis of the calculation engine, consider isolating the inputs of your optimization problem (objective, variable, and constraints) into distinct modules so that you can easily identify them.
By default, define your decision variables as Real unless you have a strong reason not to. A model built only from Real variables is a Linear Program (LP). Optimizer solves it quickly, and it is almost guaranteed to return the best possible answer.
Adding Integer or Boolean variables turns the model into a Mixed Integer Program (MIP), which requires Optimizer to test many combinations of whole-number values. Solve times can increase from seconds to minutes or hours. For larger models, the run may reach its time limit and return a good solution rather than the optimal one.
One suggestion to improve Optimizer performance is to use the Use Current Solution option. It uses the current solution to shorten the time needed for the next run, which is particularly useful when demand or data changes very little, or when you have a solid initial solution.
