Introduction to gbase 8s checkpoint

        Gbase 8s To improve transaction performance , Use buffer pool Cache data accessed by transactions , Modified the record in transaction logic , Data is not really in a transaction commit Update to disk when . In order to ensure the data integrity of the database under abnormal conditions 、 accuracy , The database uses Logical log and physical log mechanism . In normal operation mode , In order that the database can recover from exceptions in time , The database needs to be updated regularly buffer pool Synchronize your data to disk , Ensure the physical validity of the data .

         The database periodically updates buffer pool The dirty data （ Modified data ） Refresh to disk , Reach disk 、 The point in time of memory data consistency is called checkpoint . When the database recovers from an abnormal failure, it can recover from the latest checkpoint Start recovery at .Checkpoint occurs , You only need to synchronize the modified data in memory to disk , Instead of writing all the information back to disk . meanwhile , The data written back to disk is still in memory , from LRU Of M The queue moves to LRU Of F In line .

         When the system triggers Non blocking checkpoints In operation , First, block the processing of transactions , Record system recovery information , Clearing the contents of the logical log buffer , Record checkpoint information ; after , Transaction operations in the system are no longer blocked , Resume operation , The system clears all modified data in memory to disk ; Last , Record the checkpoint information in the system reservation page , This completes a non blocking checkpoint operation .