You are here

parameter

InnoDB Buffer Pool Instances is too small

If you are using MariaDB/MySQL 5.5 and newer you should use several InnoDB Buffer Pool Instances for performance reasons.
Some rules to size InnoDB Buffer Pool instances are:

  • One InnoDB Buffer Pool Instance should be at least 1 Gibyte in size (innodb_buffer_pool_size / innodb_buffer_pool_instances >= 1 Gib).
  • InnoDB Buffer Pool Instances you can set equal to the number of cores of your machine.

Table definition cache too small

The number of table definitions (SHOW CREATE TABLE\G) that can be stored in the table definition cache (table_definition_cache). If you have a large number of tables (> 400) in your database instance, you should consider a larger table definition cache to increase your database throughput and decrease your query latency.

Table open cache too small

The Table Open Cache (table_open_cache or old name table_cache) is a cache to store file handles for all threads. The actual value of cache entries can be seen with the global status of open tables (Open_tables).
Increasing table_open_cache increases the number of file descriptors (open_files_limit) that MySQL requires.

InnoDB Flush Method has changed

The InnoDB Flush Method has changed. This can have an impact on InnoDB write Performance.

MySQL Cluster Local Checkpoint (LCP) and Global Checkpoint (GCP)

MySQL Cluster is mainly an in-memory database. Nevertheless it requires a good I/O system for writing various different information to disk.

The information MySQL Cluster writes to disk are the:

  • Global Checkpoints (GCP) which are the transactions.
  • Local Checkpoints (LCP) which is a dirty image of the data.
  • Backup.

In the following schema (a 2-node Cluster) you can see what is related to each other:

lcp_gcp.png

Some more details about DiskSyncSize

The parameter DiskSyncSize is a MySQL Cluster parameter and was added in MySQL 5.1.23.

After the amount of stored bytes of data per file, the data node will fsync (flush) the LCP file to disk, even if a fsync is not needed for consistency.

This is done because the OS will otherwise buffer all the writes, and when a fsync is really needed, it can take a lot of time...

Originally this parameter was hard coded. Now it defaults to 4 Mbyte.

Subscribe to RSS - parameter