1. Why can I set it to multiple redo log? (Innodb_log_files_in_group, the default value and the recommended value are 2, we set the line of unity is 4);
2. What conditions will trigger the brush dirty? In addition to master_thread\ mandatory checkpoint, whether this frequency can be adjusted;
3. Recovery phase, whether BP is enabled. If enabled, does it occupy a particularly large amount of memory according to the MY.CNF setting?
4. Is the Redo log recovery phase parallel and can it be parallel?
5. Record format to see one, record and recovery phase;
6. Circular redo log file, how to mark the beginning and end of the ring;
7. Recovery stage read into the hash table and subsequent processing, serial or parallel;
8. Recovery stage from redo log read into the hash, one read how much, can not be unlimited read it.
9. Mtr_commit process, how to write to Log_sys->buf, is the first lock->memcpy-> release lock, or the first lock-up space-release lock->memcpy? Can there be room for optimization?
Log_sys->buf how to write to the file. and 91 blocks, which constitute the log_sys->buf of the first-in-one;
A. Redo log's group commit implementation.