InnoDB Performance Optimisation
1 like2,273 views
The document presents an overview of InnoDB performance optimization by MyDBOps, focusing on features, architecture, and performance tuning strategies. Key topics include the history of InnoDB, effective buffer pool configuration, logging mechanisms, and general recommendations for database management. MyDBOps specializes in providing expert-level support for MySQL and related technologies, emphasizing the significance of architecture knowledge for performance enhancement.
InnoDB Performance Optimisation
- 1. InnoDB Performance Optimisation Mydbops Database Meetup Presented by Karthik P R www.mydbops.com [email protected]
- 2. Mydbops is into MySQL/MongoDB Support and Consulting. It is founded by experts who have scaled database at Yahoo! ,Percona and Datavail. We are providing an expert level support and 24*7 monitoring for MySQL databases and its related technologies like MariaDB , Percona ( also clustering ) . We support modern database technologies in MySQL which includes Galera ( Clustering ), Group Replication , SQL aware Load balancers like Maxscale / ProxySQL. About Mydbops
- 3. CEO / DB Architect About Me
- 4. Agenda ● InnoDB History ● InnoDB Features ● InnoDB Architecture ● InnoDB Performance Tuning ● General Recommendations
- 5. InnoDB History ● Introduced in 1995 by Finland based company called Innobase ● In 2000, Innobase started collaboration with MySQL AB ● Oracle acquired Innobase in October 2005 ● Become Default storage engine for Mysql 5.5 in 2010 ● JSON Support introduced in MySQL 5.7 ● Separated Undo Log in MySQL 8.0
- 6. InnoDB Features ● ACID transaction Engine ● B+Tree Based Indexing ● Index Organised tables ● MVCC Support. ● Efficiently designed for OLTP workload. ● Foreign Key Support. ● Virtual column and Full text search support
- 7. InnoDB Architecture Why InnoDB Architecture ? ● Fine Tuning needs Architecture Knowledge ● Awareness about Internals ● Data Structures ( on disk / In Memory ) ● Transaction , locking . MVCC ● Checkpoint , Flushing , Purging
- 12. InnoDB Performance Tuning InnoDB Buffer Pool ● Global memory for InnoDB ● Caches the Index and data pages ● Locks and Dictionary cache ● Adaptive hash index ● Change buffering InnoDB buffer pool can be resized online in MySQL 5.7
- 13. InnoDB Performance Tuning InnoDB Buffer Pool ● LRU based Flushing ● Young and Old ( LRU ) ● O_Direct is best flushing ● Compressed and Uncompressed pages in memory
- 14. InnoDB Performance Tuning InnoDB Buffer Pool Config ● InnoDB_buffer_pool_size (70%-80% of Memory) ● innodb_buffer_pool_instances ( No of CPU core ) ● innodb_buffer_pool_dump_pct ( > 70 % is good ) ● innodb_buffer_pool_chunk_size
- 15. InnoDB Performance Tuning InnODB Change Buffer
- 16. InnoDB Performance Tuning InnoDB Change Buffer ● Present on both Disk and in memory ● Buffer pool hold Change buffer ● B+tree writes are buffered ● Restrict the size of change buffer Innodb_change_buffering=all Innodb_change_buffer_max_size ( 20-30 )
- 17. InnoDB Performance Tuning InnoDB Adaptive Hash Index ● Hash table lookup to B+tree ● Fasten the access inside buffer pool ● Works for Primary and Secondary keys ( OLTP ) ● Do not work well with text or queries with multiple join innodb_adaptive_hash_index innodb_adaptive_hash_index_parts
- 18. InnoDB Performance Tuning InnoDB Log Buffer ● Holds the data in memory before flushing to disk ● Larger log file reduce contention ● Can Flush on every commit ( for Durability ) ● Defaults work pretty well for most workloads. Innodb_log_buffer_size
- 19. InnoDB Performance Tuning InnoDB flush log at trx commit ● Controls the flushing of data from log buffer to log file. ● Supports 3 different flush behaviour ● More durability set is as default value 1 ● innodb_flush_log_at_timeout controls the flush frequency
- 20. InnoDB Performance Tuning InnoDB Flush log at trx commit
- 21. InnoDB Performance Tuning InnoDB log File Size ● Two physical log files of equal size. ● Records all changes to tablespace ● Helps in crash recovery ● Overwritten when the page is flushed from buffer pool ( Checkpoint ) ● innodb_log_files_in_group controls the number of files
- 22. InnoDB Performance Tuning InnoDB log File Size ● Two physical log files of equal size. ● Records all changes to tablespace ● Helps in crash recovery ● Overwritten when the page is flushed from buffer pool ( Checkpoint ) ● innodb_log_files_in_group controls the number of redo log files
- 23. InnoDB Performance Tuning InnoDB log File Size ● Larger log file better write performance ● Tradeoff longer recovery time ● Defaults is 48 MB is low for any production workload innodb_log_file_size=(>512M) ( Size can be calculated by innodb_os_log_written) innodb_log_files_in_group=2 Note : PMM calculates very efficiently co-relate with checkpoint
- 24. InnoDB Performance Tuning Innodb Flush method ● How data to InnoDB data files and log files is flushed ? ● default is FSYNC ● It can be numeral in MySQL 8.0 ( 0,1,2,3,4,5) innodb_flush_method=O_DIRECT(4)
- 25. InnoDB Performance Tuning Isolation Level ● How transaction isolated from each other ? ● A major role in locking along with work load ● Default Repeatable read. transaction_isolation=read_committed/repeatable_read
- 26. InnoDB Performance Tuning innodb_dedicated_server (8.0) ● Manages major tuning parameter based on your system memory ○ innodb_buffer_pool_size ○ innodb_log_file_size ○ innodb_log_file_in_group ○ innodb_flush_method ● Enable it only on dedicated MySQL server.
- 27. InnoDB Performance Tuning ● innodb_io_capacity ● innodb_io_capacity_max ● innodb_doublewrite ● innodb_checksum_algorithm ● innodb_autoinc_lock_mode ● innodb_purge_threads
- 28. InnoDB Performance Tuning ● innodb_flush_neighbors ● innodb_stats_on_metadata ● innodb_stats_persistent_sample_pages ● innodb_max_dirty_pages_pct ● innodb_max_purge_lag ● innodb_max_purge_lag_delay
- 29. InnoDB Performance Tuning ● innodb_thread_concurrency ● innodb_concurrency_tickets ● innodb_file_format ● innodb_page_size ● innodb_write_io_threads ● innodb_read_io_threads
- 30. General Recommendations ● Ensure there is primary key or Unique key in all tables. ● Smaller the primary key better they will be. ● Avoid shared tablespace for innodb tables. ● Consider partition for huge tables ( query might need changes ). ● Do not make config changes without knowing its production impact. ● 80-90% of performance gain can be achieved by tweaking 4-6 variables.
- 32. Thank You