As I understand druid can have sort of distributed segments and inverted indexes(bitmaps). Document says " inverted indexes allow for quick filtering operations (specifically, bitmaps are convenient for quickly applying AND and OR operators)." Once it finds out all the rows indexes that matches with filter criteria how exactly druid gets those data off the disk (ssd or hdd)? Is it going to do single thread random access per segment or multiple sequential search based on some grouping? Just want to understand how that pattern is optimize since you could be fetching 100s of thousands or rows if not millions per query even after filtering.