1. The buffer pool is 80% of the game

innodb_buffer_pool_size is InnoDB's in-memory cache of table and index pages.
If your working set (the data actively queried) fits in the pool, reads are
served from RAM and the server flies. If it does not, every cache miss is a disk
read.

Sizing it

On a dedicated database server, set the pool to ~60–75% of system RAM.
The remaining RAM covers the OS, per-connection buffers, and other processes.

System RAM	Reasonable buffer pool
8 GB	5 GB
16 GB	11 GB
32 GB	24 GB
64 GB	48 GB

Do not size the pool larger than your data + indexes — there is nothing to
cache beyond that. Check total InnoDB data size with sql/04_table_and_index_sizes.sql.

Verifying it is big enough

Run sql/03_innodb_status.sql. The buffer-pool hit ratio should sit at ~99%+
for a steady OLTP workload:

hit_ratio = 1 - (Innodb_buffer_pool_reads / Innodb_buffer_pool_read_requests)

Innodb_buffer_pool_reads = physical reads from disk (cache misses).
Innodb_buffer_pool_read_requests = total logical reads.

If the ratio is well below 99% and the pool is already a healthy fraction of RAM,
the fix is usually fewer pages touched (better indexes — see
02_slow_query_analysis.sql) rather than more RAM. `Innodb_buffer_pool_wait_free

0` means the pool is under genuine memory pressure.

Keep the cache warm across restarts

innodb_buffer_pool_dump_at_shutdown = ON
innodb_buffer_pool_load_at_startup  = ON

This persists the list of hot pages so a restart does not start cold with a
latency spike while the pool re-warms.

Reduce mutex contention on big machines

innodb_buffer_pool_instances = 8     # ~1 per 1-2 GB of pool, on many-core hosts

Included with MySQL Admin Scripts

Ready for the complete MySQL Admin Scripts?