Fogo's answer 2020-09-01：
checkpoint checkpoint： Batch swipe the disk .
To a certain extent, it represents the log sequence number of the log when it is flushed to disk （LSN） Location . Mark the point in the redo log that has finished brushing to disk , If there are many redo logs in the buffer pool , Full recovery requires 1 minute ,checkpoint Maybe it's marked to 58 The position of the second , At this point, the database recovery only needs to be redone. Finally 2 Data log in seconds ,checkpoint Shorten the recovery time of database .
Save it savepoint： Partial rollback .
When a transaction is rolled back, it can only be rolled back to the savepoint , There's no need to roll back to the starting point .
In the middle midpoint： Intermediate insertion .
Don't put the latest ones into LRU Header , But according to the configuration, it should be placed in a certain proportion , This place is called midpoint, midpoint The former became new list , The later became old list . At a certain time （innodb_old_blocks_time） Inside , Reading data does not old List changes new list , After this time , Read again , Meeting old List changes new list （pages made young）.innodb_old_blocks_time Set to 0, You can read it once and become new list ; Set to greater than 0 Value , You have to read it more than twice , To be new list .