Comments on: MySQL: Innodb Memory Usage Formula

By: Ole Morten Halvorsen

Ole Morten Halvorsen — Fri, 26 Dec 2008 13:50:14 +0000

(Seems like the links was stripped from my previous comment)

You could also try http://www.omh.cc/mycnf/ or http://wiki.mysqltuner.com/MySQLTuner which both will assist you in figuring out MySQL memory usage.

By: Ole Morten Halvorsen

Ole Morten Halvorsen — Fri, 26 Dec 2008 13:49:15 +0000

Hi,

You could also try or which both will assist you in figuring out MySQL memory usage.

By: Shlomi Noach

Shlomi Noach — Thu, 25 Dec 2008 03:55:54 +0000

Hi,

Missing from the equation is the thread_stack_size (per connection), query_cache_size (global).

There isn’t anything I would add to the equation which relates to InnoDB. buffer_pool_size is the only thing I consider (to be honest, I don’t even count query_cache_size as this usually measures in a few MB)

Regards

By: jon

jon — Wed, 24 Dec 2008 19:17:08 +0000

Hi Shlomi,
Yes, it’s the worst case scenario. However, I’ve seen it happen often enough to know that a system won’t recover once it starts paging to disk.

What else would you factor in, assuming only innodb is in use?

By: Shlomi Noach

Shlomi Noach — Wed, 24 Dec 2008 06:07:39 +0000

Hi,

Just a few remarks:
* The formula you’ve presented does not relate directly to InnoDB. The innodb_buffer_pool_size is in fact the only innodb related parameter there.

* It also formulates a worst case scenario where all possible connections are open, all querying, all using the sort buffer (you may as well have thrown in the join_buffer_size).

* The formula also measures key_buffer_size. If you’re using MyISAM tables, than make sure the memory sum of the formula is well below system memory, since MyISAM really needs OS page cache.

Regards