Hit ratio

The hit ratio measures how many InnoDB page reads are satisfied directly from the Buffer Pool, without touching the disk. It is the primary indicator of how effectively the memory allocated to InnoDB is being used: a high value means the active working set fits in RAM; a low value means frequent physical I/O and higher latencies.

How to read it #

MySQL reports the hit ratio inside SHOW ENGINE INNODB STATUS, under the BUFFER POOL AND MEMORY section:

Buffer pool hit rate X / 1000

A value of 997 / 1000 means 997 out of every 1000 reads were served from memory. The widely accepted operational threshold is 990/1000: dropping below it signals that the Buffer Pool is undersized relative to the active working set.

Operational context #

The hit ratio is particularly volatile after a MySQL restart: the Buffer Pool is cold and the first queries trigger cascading page faults, pushing the value well below 900/1000. InnoDB addresses this with automatic Buffer Pool dump and reload (innodb_buffer_pool_dump_at_shutdown / innodb_buffer_pool_load_at_startup), which significantly shortens the warm-up window.

A chronically low hit ratio is not always fixed by increasing innodb_buffer_pool_size alone. Inefficient queries performing full scans on large tables pollute the Buffer Pool with single-use pages, displacing hot data regardless of how much memory is available.