Coinciding with the new native data dictionary in MySQL 8.0, we have made a number of useful enhancements to our INFORMATION_SCHEMA subsystem design in MySQL 8.0. In this post I will first go through our legacy implementation as it has stood since MySQL 5.1, and then cover what’s changed.
Background
INFORMATION_SCHEMA was first introduced into MySQL 5.0, as a standards compliant way of retrieving meta data from a running MySQL server. When we look at the history of INFORMATION_SCHEMA there have been a number of complaints about the performance of certain queries, particularly in the case that there are many database objects (schemas, tables etc).
In an effort to address these reported issues, since MySQL 5.1 we have made a number of performance optimizations to speed up the execution of INFORMATION_SCHEMA queries. The optimizations are described in the MySQL manual, and apply when the user provides an explicit schema name or table name in the query.
Alas, despite these improvements INFORMATION_SCHEMA performance is still a major pain point for many of our users. The key reason behind these performance issues in the current INFORMATION_SCHEMA implementation is that INFORMATION_SCHEMA tables are implemented as temporary tables that are created on-the-fly during query execution. These temporary tables are populated via:
- Meta data from files, e.g. table definitions from .FRM files.
- Details from storage engines, e.g. dynamic table statistics.
- Data from global data structures in the MySQL server.
For a MySQL server having hundreds of database, each with hundreds of tables within them, the INFORMATION_SCHEMA query would end-up doing lot of I/O reading each individual FRM files from the file system. And it would also end-up using more CPU cycles in effort to open the table and prepare related in-memory data structures. It does attempt to use the MySQL server table cache (the system variable ‘table_definition_cache‘), however in large server instances it’s very rare to have a table cache that is large enough to accommodate all of these tables.
One can easily face the above mentioned performance issue if the optimization is not used by the INFORMATION_SCHEMA query. For example, let us consider the two queries below,
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 |
mysql> EXPLAIN SELECT TABLE_NAME FROM INFORMATION_SCHEMA.TABLES WHERE -> TABLE_SCHEMA = 'test' AND TABLE_NAME = 't1'\G *************************** 1. row *************************** id: 1 select_type: SIMPLE table: TABLES partitions: NULL type: ALL possible_keys: NULL key: TABLE_SCHEMA,TABLE_NAME key_len: NULL ref: NULL rows: NULL filtered: NULL Extra: Using where; Skip_open_table; Scanned 0 databases 1 row in set, 1 warning (0.00 sec) mysql> EXPLAIN SELECT TABLE_NAME FROM INFORMATION_SCHEMA.TABLES WHERE -> TABLE_SCHEMA like 'test%' AND TABLE_NAME like 't%'\G *************************** 1. row *************************** id: 1 select_type: SIMPLE table: TABLES partitions: NULL type: ALL possible_keys: NULL key: NULL key_len: NULL ref: NULL rows: NULL filtered: NULL Extra: Using where; Skip_open_table; Scanned all databases 1 row in set, 1 warning (0.00 sec) |
As we can see from the EXPLAIN output, we see that the former query would use the values provided in WHERE clause for the TABLE_SCHEMA and TABLE_NAME field as a key to read the desired FRM files from the file system. However, the latter query would end up reading all the FRM in the entire data directory, which is very costly and does not scale.
Changes in MySQL 8.0
One of the major changes in 8.0 is the introduction of a native data dictionary based on InnoDB. This change has enabled us to get rid of file-based metadata store (FRM files) and also help MySQL to move towards supporting transactional DDL. For more details on introduction of data dictionary feature in 8.0 and its benefits, please look at Staale’s post here.
Now that the metadata of all database tables is stored in transactional data dictionary tables, it enables us to design an INFORMATION_SCHEMA table as a database VIEW over the data dictionary tables. This eliminates costs such as the creation of temporary tables for each INFORMATION_SCHEMA query during execution on-the-fly, and also scanning file-system directories to find FRM files. It is also now possible to utilize the full power of the MySQL optimizer to prepare better query execution plans using indexes on data dictionary tables.
The following diagram explains the difference in design in MySQL 5.7 and 8.0.
If we consider the above example under Background, we see that the optimizer plans to use indexes on data dictionary tables, in both the cases.
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 |
mysql> EXPLAIN SELECT TABLE_NAME FROM INFORMATION_SCHEMA.TABLES WHERE TABLE_SCHEMA = 'test' AND TABLE_NAME = 't1'; +--+-----------+-----++------+------------------+----------++----------------------+----+--------+----------------------------------+ |id|select_type|table||type |possible_keys |key ||ref |rows|filtered|Extra | +--+-----------+-----++------+------------------+----------++----------------------+----+--------+----------------------------------+ | 1|SIMPLE |cat ||index |PRIMARY |name ||NULL | 1| 100.00|Using index | | 1|SIMPLE |sch ||eq_ref|PRIMARY,catalog_id|catalog_id||mysql.cat.id,const | 1| 100.00|Using index | | 1|SIMPLE |tbl ||eq_ref|schema_id |schema_id ||mysql.sch.id,const | 1| 10.00|Using index condition; Using where| | 1|SIMPLE |col ||eq_ref|PRIMARY |PRIMARY ||mysql.tbl.collation_id| 1| 100.00|Using index | +--+-----------+-----++------+------------------+----------++----------------------+----+--------+----------------------------------+ mysql> EXPLAIN SELECT TABLE_NAME FROM INFORMATION_SCHEMA.TABLES WHERE TABLE_SCHEMA like 'test%' AND TABLE_NAME like 't%'; +--+-----------+-----++------+------------------+----------++-----------------------+----+--------+---------------------------------+ |id|select_type|table||type |possible_keys |key || ref |rows|filtered|Extra | +--+-----------+-----++------+------------------+----------++-----------------------+----+--------+---------------------------------+ | 1|SIMPLE |cat ||index |PRIMARY |name || NULL | 1| 100.00|Using index | | 1|SIMPLE |sch ||ref |PRIMARY,catalog_id|catalog_id|| mysql.cat.id | 6| 16.67|Using where; Using index | | 1|SIMPLE |tbl ||ref |schema_id |schema_id || mysql.sch.id | 26| 1.11|Using index condition;Using where| | 1|SIMPLE |col ||eq_ref|PRIMARY |PRIMARY || mysql.tbl.collation_id| 1| 100.00|Using index | +--+-----------+-----++------+------------------+----------++-----------------------+----+--------+---------------------------------+ |
When we look at performance gain with this new INFORMATION_SCHEMA design in 8.0, we see that it is much more efficient than MySQL 5.7. As an example, this query is now ~100 times faster (with 100 databases with 50 tables each). A separate blog will describe more about performance of INFORMATION_SCHEMA in 8.0.
|
1 2 3 |
SELECT TABLE_SCHEMA, TABLE_NAME, TABLE_TYPE, ENGINE, ROW_FORMAT FROM information_schema.tables WHERE TABLE_SCHEMA LIKE 'db%'; |
Sources of Metadata
Not all the INFORMATION_SCHEMA tables are implemented as a VIEW over the data dictionary tables in 8.0. Currently we have the following INFORMATION_SCHEMA tables designed as views:
- SCHEMATA
- TABLES
- COLUMNS
- VIEWS
- CHARACTER_SETS
- COLLATIONS
- COLLATION_CHARACTER_SET_APPLICABILITY
- STATISTICS
- KEY_COLUMN_USAGE
- TABLE_CONSTRAINTS
Upcoming MySQL 8.0 versions aims to provide even the following INFORMATION_SCHEMA tables as views:
- EVENTS
- TRIGGERS
- ROUTINES
- REFERENTIAL_CONSTRAINTS
To describe the INFORMATION_SCHEMA queries which are not directly implemented as VIEWs over data dictionary tables, let me first describe that there are two types of meta data which are presented in INFORMATION_SCHEMA tables:
- Static table metadata. For example: TABLE_SCHEMA, TABLE_NAME, TABLE_TYPE, ENGINE. These statistics will be read directly from the data dictionary.
- Dynamic table metadata. For example: AUTO_INCREMENT, AVG_ROW_LENGTH, DATA_FREE. Dynamic metadata frequently changes (for example: the auto_increment value will advance after each insert).In many cases the dynamic metadata will also incur some cost to accurately calculate on demand, and accuracy may not be beneficial for the typical query. Consider the case of the DATA_FREE statistic which shows the number of free bytes in a table – a cached value is usually sufficient.
In MySQL 8.0 (Prior to 8.0.3 RC1), the dynamic table metadata will default to being cached. This is configurable via the setting information_schema_stats (default cached), and can be changed to information_schema_stats=latest in order to always retrieve the dynamic information directly from the storage engine (at the cost of slightly higher query execution).
As an alternative, the user can also execute ANALYZE TABLE on the table, to update the cached dynamic statistics.
In MySQL 8.0.3 RC1 onwards we have replaced the option information_schema_stats with a new option information_schema_stats_expiry . Please see Further improvements on INFORMATION_SCHEMA in MySQL 8.0.3 for details.
Conclusion
The INFORMATION_SCHEMA design in 8.0 is a step forward enabling:
-
- Simple and maintainable implementation.
- Us to get rid of numerous INFORMATION_SCHEMA legacy bugs.
- Proper use of the MySQL optimizer for INFORMATION_SCHEMA queries.
- INFORMATION_SCHEMA queries to execute ~100 times faster, compared to 5.7, when retrieving static table metadata, as show in query Q1.
There is more to discuss about INFORMATION_SCHEMA in 8.0. The new implementation comes with a few changes in behavior when compared to the old INFORMATION_SCHEMA implementation. Please check the MySQL manual for more details about it.
Thanks for using MySQL!