in Rdb_iterator_base::next_with_direction, it checks prefix and lower bound and upper bound, this consuming too many cpu in our cases, when running tpcc, the flame graph showing these 3 compare takes up to 3% cpu time, which is ~10% of myrocks::Rdb_iterator_base::next_with_direction.

MyRocks use [Reverse]BytewiseComparator, so the optimal should be just 1 compare in scanning, I have tried to improve this way, it is ok in most cases but failed some mtr tests. I am expecting an official fix for this issue.