Cluster Computing
Optimization always involves several iterations, and it is normal that the forward problem has to be solved 100–1000 times before a good solution is found. Therefore it is common that computational cost becomes important. To address this cost, you can use some of these methods:
•
Reduce the mesh size (while still capturing the relevant physical effects)
•
Exploit symmetry
•
Eliminate couplings and physical effects that are of lesser importance
•
Make sure that the solver setup is efficient
The computational cost might still be higher after following these best practices. Cluster computing can be used to reduce the cost further. There are three options (see below), and they cannot be combined. The first option is used by default. The settings for the other options are accessible via the
Batch and Cluster
checkbox in the
Show More Options
dialog.