"One method would be to break up data into multiple slaves. This
could be done database by database (schema by schema) creating one
slave database. Each slave would be pointing to the same master,
and using replicate_do_db and replicate_ignore_db statements to
specify which data to capture for it's slave. However it may be
that your applications are not broken up cleanly that way, or that
load favors certain tables, rather than certain schemas. So for
your site there may be reason to do it on a table-by-table basis.
You would then use replicate_do_table and replicate_ignore_table.
You'll also need to keep in mind that more slaves mean more threads
on the master database, which could add load there. If you run into
that problem, consider creating a distribution master, and then
slaving off of that. As a further consideration, realize that if
you are running backups off of slave databases, your backup scripts
will obviously be more complicated in this arrangement, so document
well."
