It was suggested @2019 that the documentation will get fixed soon . Also we are testing a 20 million insert 50 million kafka upsert and 1 big aggregation join query at the same time . We will like know which one to use in this case . MMAP or HEAP ? But given the current state of documentation we need to check in the codebase . Any guidance is welcome .

I think that discussion may not apply in the current state of docs (need to go over it). But the guidance is to always use load-mode as MMAP.

Thanks @Mayank

Load mode applies to how the immutable segments are loaded. MMAP - they are memory mapped. HEAP - they are loaded in direct memory (off-heap, so the naming is incorrect).