MySQL AB excels at rapidly fixing issues and introducing new features to the entire product line. Many of these new features have a significant impact on performance. Unfortunately, for the average overworked, underpaid database developer or administrator, it can be difficult to keep up with all of these new capabilities. In fact, in some cases it's likely that beneficial upgrades are put off because the administrator is unaware of the advantages of upgrading.
Because this book focuses on advancing MySQL performance, some of the major database speed augmentations provided in MySQL versions beginning with 4.0 are listed. These product enrichments show a clear pattern of continual performance-related improvements over time.
For brevity's sake, other enhancements that don't really impact system response are omitted. Internal engine improvements and bug fixes are also skipped, unless they provide direct, controllable access to developers. Finally, note that each of the topics listed in Table 3.1 are covered in the appropriate chapter.
Table 3.1. MySQL Performance-Related Features by Version
This feature provides a fast interface to MyISAM tables, letting developers directly position and move anywhere within the table, and then operate on rows accordingly.
Several server variables give administrators more control over how FULLTEXT indexes are built and managed.
Improved SQL capability also yields potentially better index utilization.
Two new features (SQL_CALC_FOUND_ROWS, FOUND_ROWS()) give developers better visibility into costs and expected results from a query.
You can cache queries and their results, which adds value to your applications regardless of your chosen storage engine.
This feature introduces the innodb_thread_concurrency server setting, improving the speed of parallel operations for the InnoDB storage engine.
Track long queries
This feature adds the long_query_time configuration setting to provide better log tracking of problem queries. Queries that exceed this threshold are logged.
Improved ORDER BY
Indexes are now more efficiently used for additional sorting scenarios.
The added read_buffer_size setting gives administrators more control over sequential read performance.
This feature reduces the amount of index creation work performed by LOAD DATA INFILE if data is already present in table.
Four transaction isolation levels are now available for developers to control concurrency.
InnoDB now has improved deadlock avoidance and detection algorithms.
Query cache details
The Qcache_lowmem_prunes indicator tells administrators how often contents of the query cache had to be removed because of insufficient memory.
Avoid table scans
Developers can now use the FORCE INDEX syntax to override the optimizer's query plan, thus going even further to avoid a costly table scan.
Administrators can now define their own lists of words to be ignored in FULLTEXT searches via the --ft-stopword-file option.
MyISAM and threading
New parallel table repair and index creation features (configured via new myisam_repair_threads setting) introduce potential for significant index creation speed improvements.
InnoDB buffer pool
You can now specify how many pages in the InnoDB buffer pool are allowed to be dirty (that is, have altered data or index information) by setting innodb_max_dirty_pages_pct.
Limit thread delay
The new max_delayed_threads variable controls how many threads are allowed to queue to perform their inserts.
The max_seeks_for_key setting helps drive the optimizer toward choosing an index-based query plan, even if the index holds very duplicate information.
The --read-only parameter for mysqld prevents inadvertent writes to a slave server.
Five new server variables let administrators more accurately allocate buffer memory. Variables include TRansaction_alloc_block_size, TRansaction_prealloc_size, range_alloc_block_size, query_alloc_block_size, and query_prealloc_size.
Administrators can now tune the new innodb_table_locks session variable to reduce the likelihood of deadlocks.
MEMORY B-tree index
Database designers can now elect to use a B-tree index on a MEMORY table, instead of the default hash index.
MySQL now supports the extended (up to 64GB) memory capabilities (AWE) on Windows servers.
The EXPLAIN query report now provides additional data useful in helping determine if a query is as efficient as possible.
Running mysqldump now disables foreign key checks automatically when generating a load file, helping to speed the reloading process.
Several new important features for the MyISAM engine are now available. First, administrators can now use symbolic links for MyISAM tables, which lets tables be spread among multiple disk drives if desired.
| || |
Next, key cache performance has been boosted by allowing for midpoint insertions, as well as permitting multiple threads to simultaneously access the cache.
Temp round robin
Administrators can now configure several directories to serve as temporary storage for MySQL by setting the tmpdir parameter. This can help to balance the disk load among multiple drives.
You can now create multiple, specialized instances of the MyISAM performance-enhancing key cache. The new preload_buffer_size setting lets administrators configure memory when preloading indexes.
The new max_length_for_sort_data setting helps MySQL determine what kind of file sort algorithm to use when processing an ORDER BY.
You can now specify up to 1,000 bytes for a MyISAM table's index key; you can create up to 64 indexes per table for InnoDB and MyISAM.
Large table support
You can now set MyISAM's row pointer size (myisam_data_pointer_size), which lets you address very large tables.
The new innodb_autoextend_increment setting lets you control much additional disk space InnoDB requests when growing a tablespace.
The innodb_max_purge_lag setting lets you control what happens when there is a significant amount of information to purge from internal InnoDB logs.
New parameters now let you use MySQL to generate a point-in-time InnoDB backup.
The MySQL optimizer is now able to create query plans that use multiple indexes to satisfy an OR clause.
You can now create server-side stored procedures, helping to remove workload from clients as well as centralize software development.
Views provide numerous benefits for administrators, including letting them define relationships among multiple tables, specify filter criteria, and present a simpler data interface to developers.
The optimizer_prune_level and optimizer_search_depth settings let you dictate how you want the MySQL optimizer to examine potential query plans. Also, you can use the new Last_query_cost indicator to get an idea of the price of your most recent query.
Many new indicators for InnoDB report on its current status and workload.
The updateable_views_with_limit setting helps prevent a runaway update of a view when LIMIT is specified.
This feature lets you define structures to hold retrieved information. Typically used in conjunction with stored procedures, they add significant processing flexibility to your applications.
Triggers let you define data-driven events that will kick off activities on the server.
This feature defines table structures and operates upon data in remote locations as if they were local.