MySQL InnoDB performance improvement: InnoDB buffer pool instances – Updated!

Are you running into MySQL load problems? Learn how to tune MySQL servers for a heavy InnoDB workload, by configuring innodb_buffer_pool_instances. Dividing the InnoDB buffer pool into multiple instances improves Disk I/O. By doing so, you run your database and website more efficiently and faster. Here is a little help for you.

Tune MySQL InnoDB buffer pool instances & size for heavy workloads

All for more InnoDB Disk I/O performance on MySQL 5.5+.

Tuning MySQL servers is an ever ongoing process. Every new MySQL version brings new configuration settings you can use to improve its performance. As a MySQL DBA you want your database server and databases to perform better than well, don’t you?

MariaDB/MySQL 5.5.4 introduces new configuration settings for the InnoDB storage engine. This can greatly improve MySQL’s InnoDB performance, both in read and write operations.

One of those settings is innodb_buffer_pool_instances. The innodb_buffer_pool_instances divides the InnoDB buffer pool into separate instances. Dividing your buffer pool into separate instances can improve concurrency, by reducing contention as different threads read and write to cached pages. Multiple buffer pool instances are configured using the innodb_buffer_pool_instances configuration option.

You might also want to adjust the innodb_buffer_pool_size value:

The larger the InnoDB buffer pool, the more InnoDB acts like an in-memory database. It reads data from disk once and then accesses the data from memory during subsequent reads. Buffer pool size is configured using the innodb_buffer_pool_size configuration option.

Back to increasing innodb_buffer_pool_instances:

The innodb_buffer_pool_instances divides the InnoDB buffer pool in a number of regions.

The number of regions that the InnoDB buffer pool is divided into. For systems with buffer pools in the multi-gigabyte range, dividing the buffer pool into separate instances can improve concurrency, by reducing contention as different threads read and write to cached pages. Each page that is stored in or read from the buffer pool is assigned to one of the buffer pool instances randomly, using a hashing function. Each buffer pool manages its own free lists, flush lists, LRUs, and all other data structures connected to a buffer pool, and is protected by its own buffer pool mutex.

This option takes effect only when you set the innodb_buffer_pool_size to a size of 1 gigabyte or more. The total size you specify is divided among all the buffer pools. For best efficiency, specify a combination of innodb_buffer_pool_instances and innodb_buffer_pool_size so that each buffer pool instance is at least 1 gigabyte.

In MySQL versions prior to 5.5.4 this was not configurable and thus set to just one instance. Now you can increase innodb_buffer_pool_size, and you can divide the InnoDB buffer pool into multiple regions by setting innodb_buffer_pool_instances to 2, 3, 4 or 8. As long as innodb_buffer_pool_size is set high enough, and you have enough memory available in your MySQL database server. This increases InnoDB read/write threads.

Read this too:   Restore single MySQL table from a full mysqldump backup file

To enable multiple buffer pool instances, set the innodb_buffer_pool_instances configuration option to a value greater than 1 (the default) up to 64 (the maximum).

For example, you can set innodb_buffer_pool_size to 6 GB and innodb_buffer_pool_instances to 4 in your my.cnf MySQL configuration file:

; InnoDB buffer pool size in bytes. The primary value to adjust on a database server, 
; can be set up to 80% of the total memory in these environments
innodb_buffer_pool_size = 6000M

;  If innodb_buffer_pool_size is set to more than 1GB, innodb_buffer_pool_instances

; divides the InnoDB buffer pool into this many instances.
innodb_buffer_pool_instances = 4

In this example, I’ve used an innodb_buffer_pool_size of 6000M (6 GB), so there is 1500M available per innodb_buffer_pool_instance, which is more than the minimum 1 GB. As a rule of thumb, set your innodb_buffer_pool_size to approximately 70 – 80% of the RAM available.

Innodb_buffer_pool_instances defaults

Various MySQL versions have different innodb_buffer_pool_instances default values, here is an overview – listing – for you:

MySQL version# InnoDB buffer pool instancesNotes
MySQL 5.5 (<= 5.5.4)1not configurable
MySQL 5.51
MySQL 5.6 (<= 5.6.5)1
MySQL 5.6 (>= 5.6.6)8or 1 if innodb_buffer_pool_size < 1GB
MySQL 5.78or 1 if innodb_buffer_pool_size < 1GB
MariaDB 10 (<= MariaDB 10.0.3)1
MariaDB 10 (>= MariaDB 10.0.4)8

InnoDB read and write I/O threads in MySQL

Besides innodb_buffer_pool_instances, you can also increase the number of InnoDB read I/O threads and write I/O threads. These are configured with innodb_write_io_threads and innodb_read_io_threads.

Both settings default to 4 threads. We can increase these to, for example, 8:

; Number of I/O threads for writes
innodb_write_io_threads = 8
; Number of I/O threads for reads
innodb_read_io_threads = 8

innodb_read_io_threads

The number of I/O threads for read operations in InnoDB. The default value is 4.

innodb_write_io_threads

The number of I/O threads for write operations in InnoDB. The default value is 4.

When should you increase the number of innodb_read_io_threads? When you see more than 64 × innodb_read_io_threads pending read requests in SHOW ENGINE INNODB STATUS, you might gain by increasing the value of innodb_read_io_threads.

Optimizing InnoDB Disk I/O

If you follow the best practices for database design and the tuning techniques for SQL operations, but your database is still slowed by heavy disk I/O activity, explore these low-level techniques related to disk I/O. If the Unix top tool or the Windows Task Manager shows that the CPU usage percentage with your workload is less than 70%, your workload is probably disk-bound, Optimizing InnoDB Disk I/O.

Read this too:   How to test SMTP Authentication and StartTLS

Starting from MariaDB 10.0, the default number of innodb_buffer_pool_instances is 8. This means you have to configure your innodb_buffer_pool_size to at least 8 GB, see the defaults above.

Protip, don’t over optimize: never make too many configuration changes at once. After changing one or two settings, let the server run for a few days so you can learn the implications of the changes. Then, if necessary, make additional changes to the configuration.

Convert MyISAM to InnoDB tables for WordPress using a plugin
For WordPress, I created a plugin to convert MyISAM tables to InnoDB, that now is incorporated into the Vevida Optimizer WordPress plugin. The Vevida Optimizer plugin extends the automatic update feature already present in WordPress. WordPress core updates can be switched on or off, themes and translations can be automatically updated, and the plugin updates can be configured on a per-plugin basis.

Extra tips for MySQL performance tuning

Besides optimizing InnoDB for a high-performance workload, there is more you can do to tune MySQL server and database performance. Here are some extra MySQL configuration tips for you.

Some information might be outdated and obsolete but may hold valuable information for tuning your MySQL server.

Note: this is a translation and rewrite of my older Dutch post “MySQL performance en optimalisatie tips“, which is now deleted and links to here. Just in case you were wondering why you arrived here instead of the Dutch post after clicking a link :-)

1: No two MySQL servers are the same

When optimizing MySQL database servers, keep in mind that no server is equal to another. Settings that work well on one server, may degrade performance on a second. If you manage multiple servers with its configuration under version control (e.g almost -or exactly- the same MySQL configuration for all servers), choose what works best on all servers.

To determine what you can improve, you first need to know how the server performs now. You can use some MySQL commands for this on your MySQL cli (data comes from my very old post).

mysql> SHOW STATUS LIKE '%key_read%';
+-------------------+-------------+
| Variable_name     | Value       |
+-------------------+-------------+
| Key_read_requests | 11810240259 |
| Key_reads         | 9260357     |
+-------------------+-------------+

These two variables and values relate to the configured key_buffer_size

In this old example, the database server has 4 GB of RAM and a configured key_buffer_size of 512 MB. The ratio (Key_read_requests / Key_reads) is approximately 1/1275, which is good but the key_buffer_size value may be increased to 768 MB. Even though this is not yet necessary.

mysql> SHOW STATUS LIKE 'thread%';
+-------------------+---------+
| Variable_name     | Value   |
+-------------------+---------+
| Threads_cached    | 0       |
| Threads_connected | 76      |
| Threads_created   | 6234040 |
| Threads_running   | 2       |
+-------------------+---------+

These Threads_* variable values show you there are currently 76 connected threads, of which only 2 are really running a thread (executing a statement). This means 74 connections are idle.

Read this too:   Convert decimal to hex in Bash?

Here you can also see that there is no “thread cache” set up for MySQL: Threads_cached | 0

You can use the MySQL Server System variable thread_cache_size to configure how many threads must be cached by MySQL. This is one of those configuration settings that, probably, provides the least performance gain, but still…

Don’t set this one too high, somewhere between 20 and 40 is often good enough:

thread_cache_size = 20

When you execute the previous statement again, the values will be:

mysql> SHOW STATUS LIKE 'thread%';
+-------------------+-------+
| Variable_name     | Value |
+-------------------+-------+
| Threads_cached    | 14    |
| Threads_connected | 98    |
| Threads_created   | 2896  |
| Threads_running   | 1     |
+-------------------+-------+

You now have 14 threads cached :)

2: Miscellaneous MySQL configuration settings

A few words on some miscellaneous configuration settings.

2.1: tmp_table_size and max_heap_table_size
The default tmp_table_size and max_heap_table_size values are 16M. These two have to be equal in size! It sets the maximum size for internal in-memory tables, resulting in less creation of temporarily MyISAM tables on the file system. That in return, results in less disk I/O.

2.2: join_buffer_size
The join_buffer_size sets a maximum buffer size for plain index scans, range index scans and joins without indices (and therefore perform full table scans). Keep this one low, 1M for example.

3. Use Diagnostics for improvements

It is important to frequently run diagnostics and/or look up diagnostic data (for example in your information_scheme table). Percona has a lot of information about some key metrics:

MySQL tuning, the conclusion

Tuning MySQL and the InnoDB storage engine is an important step in further optimizing your hosting environment. Every new MySQL version brings new settings to improve your MySQL configuration, so be sure to read those changelogs.

In this article we went over InnoDB Buffer Pool Size and InnoDB Buffer Pool Instances. Setting these properly greatly improves your MySQL server’s performance!

But never (ever, ever) over-optimize! Please don’t make too many configuration changes at once. Make one or two and restart mysqld. After monitoring your system for a few days, running with the new configuration, you have data available to further optimize other MySQL settings.

With InnoDB being the default storage engine, you also have to make sure you make use of this storage engine in MySQL. Therefore it is important to convert old MyISAM tables to InnoDB.

21,027 views

4 comments

  1. The use of Query Cache was justified when using MyISAM tables, which were completely locked during update/insert/delete data. When using InnoDB tables on the highload progect, the Query Cache on the contrary, only interferes and dramatically degrades server performance. Therefore, Query Cache is disabled by default in new versions of the MySQL Server.

    1. Hi Dmitry, thank you for your comment.

      I thought I removed all references to query_cache. Nowadays Query Cache is removed from MySQL, because it’s only for the MyISAM table engine and not for InnoDB. Unfortunately I forgot one, and that’s gone now.

  2. I understand you used “courier” font for the configuration parameters so that they will display differently and be obvious that it’s a “computer term”, but you also need to be careful about the size of that font. As it is now, the “computer terms” are HALF the size of the normal text! It makes reading your post quite a bit more difficult. It’s like walking down a road with holes to trip in every few feet.

Leave a Reply

Your email address will not be published. Required fields are marked *