- 7 Performance and Scalability Enhancements
- 7.1 Overview of InnoDB Performance
- 7.2 Faster Locking for Improved Scalability
- 7.3 Using Operating System Memory Allocators
- 7.4 Controlling InnoDB Change Buffering
- 7.5 Controlling Adaptive Hash Indexing
- 7.6 Changes Regarding Thread Concurrency
- 7.7 Changes in the Read-Ahead Algorithm
- 7.8 Multiple Background I/O Threads
- 7.9 Asynchronous I/O on Linux
- 7.10 Group Commit
- 7.11 Controlling the Master Thread I/O Rate
- 7.12 Controlling the Flushing Rate of Dirty Pages
- 7.13 Using the PAUSE instruction in InnoDB spin loops
- 7.14 Control over Spin Lock Polling
- 7.15 Making Buffer Cache Scan Resistant
- 7.16 Improvements to Crash Recovery Performance
- 7.17 Integration with MySQL PERFORMANCE_SCHEMA
- 7.18 Improvements to Performance from Multiple Buffer Pools
- 7.19 Better Scalability with Multiple Rollback Segments
- 7.20 Better Scalability with Improved Purge Scheduling
- 7.21 Improved Log Sys Mutex
- 7.22 Separate Flush List Mutex
Starting in InnoDB storage engine 1.1, the asynchronous I/O capability
that has been supported on Windows systems, is now available on
Linux systems. (Other Unix-like systems continue to use
synchronous I/O calls.) This feature improves the scalability of
heavily I/O-bound systems, which typically show many pending
reads/writes in the output of the command show engine
innodb status\G.
If a problem with the asynchronous I/O subsystem in the OS
prevents InnoDB from starting, set the option
innodb_use_native_aio=0 in the configuration
file. This new configuration option applies to Linux systems only,
and cannot be changed once the server is running.
This is the User’s Guide for InnoDB storage engine 1.1 for MySQL 5.5, generated on 2010-04-13 (revision: 19994) .
