Вот тут Admin Zone Forums - MySQL Server Tweaking Basic обсуждают настройку MySQL, сохраню ка, я это себе, там есть что почитать.
А вот тут приводят волешбный конфиг, его тоже надо изучить

Magical config for high performance

[client]
#password = your_password
port = 3306
socket = /tmp/mysql.sock

[server]
port=3306
socket=/tmp/mysql.sock
user=mysql
datadir=/var/db/mysql/
basedir=/usr/local
max_connections=3000
max_connect_errors=10
table_cache=2048
max_allowed_packet=1M
binlog_cache_size=1M
max_heap_table_size=64M
sort_buffer_size=64K
join_buffer_size=1M
thread_cache=16
thread_concurrency=16
thread_stack=196K
query_cache_size=0
ft_min_word_len=4
default_table_type=MYISAM
transaction_isolation=REPEATABLE-READ
tmp_table_size=64M
skip-locking
server-id=1
innodb_status_file=0
innodb_data_home_dir=/var/db/mysql/
innodb_data_file_path=ibdata1:100M:autoe
xtend
innodb_log_group_home_dir=/var/db/mysql/
innodb_buffer_pool_size=500M
innodb_additional_mem_pool_size=20M
innodb_log_file_size=900M
innodb_log_files_in_group=2
innodb_log_buffer_size=8M
innodb_flush_log_at_trx_commit=1
innodb_lock_wait_timeout=300
innodb_locks_unsafe_for_binlog=1
innodb_thread_concurrency=0

Попробовал - хорошая штука. Свою такую за $218 не написать конечно.
Ещё пробовал перловые innotop-1.6.0 и mytop-1.6
Как говорится - “… жалкое подобие левой руки”, зато бесплатно.

The good news is that SQL Relay’s performance here is comparable to PHP’s persistent connections. The reason is because SQL Relay has already established X number of Oracle (or MySQL) connections and strives to maintain that number continually. Then, your PHP app connects and disconnects from the SQL Relay server, which is lightweight.

Getting everything to run smoothly does take a fair bit of tuning and careful management of your SQL Relay connections. If you have leaky code that doesn’t disconnect properly, you can still flood your connection pool, no pun intended. Wish I had benchmarks but, the main thing is that SQL Relay offers similar performance and won’t flood your database with connections, both of which are good things.

Выглядит заменчиво, надо посмотреть - вероятно это дешевая альтернатива domain data separation при больших нагрузках

1) Increase innodb_buffer_pool_size to 80% of memory of dedicated host. At least give it 265M. Here under is a snippet from mysql.ini

set-variable=innodb_buffer_pool_size=256M
set-variable=innodb_additional_mem_pool_size=20M

Without any further optimization it gives you 50% performance gain (against old setting of 32Mb)
2) Increase innodb_log_file_size to 25% of innodb_buffer_pool_size

set-variable=innodb_log_file_size=64M
set-variable=innodb_log_buffer_size=8M

Packaged with optimization above it can shave you up to additional 10%
3) Make sure that for read-only transactions your switched ON auto commit setting. And then switched it back for read-write tasks. Set read-only settings properly to JDBC statement as well.
Here you will see drastic improvement if you treated read-only transaction wrongly i.e. committed and rolled back them yourself from DAO
4) Make sure that you execute operation with DB concurrently. I was able to get 30% improvement when running 8 threads. There is a catch here though – by default internally MySQL holds 8 threads to serve DB requests. Make number of threads here and there consistent.
5) Change isolation level to READ-UNCOMMITTED if it possible (default for InnoDB is REPEATABLE READ)

set-variable=transaction-isolation=READ-UNCOMMITTED

My tests show that gain here is almost negligible around 4%