If we can somehow get the memory use per SingleBenchmark, we can add a summary graph in the benchmark report section "performance" that shows the memory use versus the problem scale. The field will love that to answer their "sizing" questions.

Can we somehow get the memory use per SingleBenchmark?
Not really. Not non-intrusively at least (= so this means it can't be activated by default).

• We can take the highest usedMemory of the ProblemStatisticType MEMORY_USE (presuming that's active). This one polls about once per second (Check this with debugger!). Do not confuse the highest usedMemory (the top spike over a period of time of the usedMemory) with the maximumMemory (as in -Xmx1024m).

If problem statistic MEMORY_USE is not active, the summary statistic should not be there either (or better yet the summary tab should be disabled in the report, so people know it exists).

If paralel benchmark count > 1, the summary statistic should be accompanied by the same warning as the problem statistic MEMORY_USE now gets in that case.

The size of the dataset can heavily affect the memory usage. Also visualize ProblemBenchmarkResult.averageUsedMemoryAfterInputSolution so they can see who's using all that memory (the dataset itself or optaplanner). Test with MachineReassingment example dataset B7 and B10.