I'm building a scheduling model with OPL in IBM ILOG CPLEX Optimization Studio v12.5.0. As I try to communicate with another program that generates the data, I use SheetRead to Excel (xlsm) and SheetWrite to write the decisions to Excel. I'm using a 8 GB RAM machine.
The model is working fine, but when I increase the puzzle size, I get performance issues. For one example, with a timelimit for solving on 200 seconds, the run configuration is finished after more than an hour. I tried a couple of things like turning of the SheetWrite and even the objective, but this doesn't seem to help. The Excel for this problem consists of 12 tabs, each having between 1 and 8 columns. The number of rows does vary heavily, between 6 and 3600. (The biggest one I have is 4 columns x 82000 lines but I'm not even using that one yet. A test with adding a sheet of 4 columns and 12000 rows resulted in one minute extra for total time, not solving time.) For writing I only have two sheets where each 4 columns are filled with respectively 3600 and 1350 rows.
When checking the (Windows) task manager, it looks like Java is the bottleneck: this has a big peak and high CPU score for a long time, OPL only runs shortly, for output both Excel and Java are working. Does anyone of you have experience with something like this?