Github user sihuazhou commented on the issue: https://github.com/apache/flink/pull/5979 @StephanEwen , I had a micro-benchmark, here is the result ``` ---------> Batch VS Put <------------ BATCH: end insert - duration:255 PUT: end insert - duration:545 ---------> MapState#Clear New VS Old <------------ ----> NEW: end delete 50 records - duration:1077719 OLD: end delete 50 records - duration:10949887 ----> NEW: end delete 100 records - duration:809182 OLD: end delete 100 records - duration:1617317 ----> NEW: end delete 200 records - duration:1970156 OLD: end delete 200 records - duration:2731749 ----> NEW: end delete 400 records - duration:2492822 OLD: end delete 400 records - duration:13894767 ----> NEW: end delete 800 records - duration:5816919 OLD: end delete 800 records - duration:13017591 ``` I tested two things: - To compare the performance between `WriteBatch()` VS `Put()` - To compare the performance of the `RocsDBMapState#Clear()`: new version vs old version In general, the more records there, the more lift we would get from the new version, here is my code for the test: https://github.com/sihuazhou/flink/commit/75504ad6fdb33755cccef43935e007bd5804ea9d
---