摘要:1 innodb_buffer_pool_instances可以開啟多個內存緩沖池,把需要緩沖的數據hash到不同的緩沖池中,這樣可以並行的內存讀寫。
2 innodb_buffer_pool_instances 參數顯著的影響測試結果,特別是非常高的 I/O 負載時。
3 實驗環境下, innodb_buffer_pool_instances=8 在很小的 buffer_pool 大小時有很大的不同,而使用大的 buffer_pool 時,innodb_buffer_pool_instances=1 的表現最棒。
1 定義
The number of regions that the InnoDB
buffer pool is divided into. For systems with buffer pools in the multi-gigabyte range, dividing the buffer pool into separate instances can improve concurrency, by reducing contention as different threads read and write to cached pages. Each page that is stored in or read from the buffer pool is assigned to one of the buffer pool instances randomly, using a hashing function. Each buffer pool manages its own free lists, flush lists, LRUs, and all other data structures connected to a buffer pool, and is protected by its own buffer pool mutex.
This option takes effect only when you set the innodb_buffer_pool_size
to a size of 1 gigabyte or more. The total size you specify is divided among all the buffer pools. For best efficiency, specify a combination ofinnodb_buffer_pool_instances
and innodb_buffer_pool_size
so that each buffer pool instance is at least 1 gigabyte.
- 測試日期: Oct-2012
- 測試目的: 測試 MySQL 5.6.7 的表現
- 硬件換
- 服務器: Dell PowerEdge R710
- CPU: 2x Intel(R) Xeon(R) CPU E5-2660 0 @ 2.20GHz
- 內存: 192GB(這個內存太猛了)
- 存儲: Very Fast PCIe Flash Card
- 文件系統: ext4
- 軟件
- 操作系統: CentOS 6.3
- MySQL 版本: 5.6.7-RC
- 測試規范
- 測試工具: tpcc-mysql
- 測試數據: 2500W (~250GB of data)
- 測試時間: 總共測試 4000 秒,但只取最后的 2000 秒,避免因為冷啟動的問題導致測試結果不准確
- 不同的測試參數: 使用幾組不同的 innodb_buffer_pool_size:13, 25, 50, 75, 100, 125GB ,innodb_buffer_pool_instances: 1 and 8, and innodb_log_file_size: 2x4GB and 2x8GB.
測試結果:
第一個結果使用的事 2x4GB 的 InnoDB 日志文件:

我們可看出當 innodb_buffer_pool_instances=8 在很小的 buffer_pool 大小時有很大的不同,而使用大的 buffer_pool 時,innodb_buffer_pool_instances=1 的表現最棒。
測試結果在大的 buffer_pool 時是很穩定的,原因是 InnoDB 使用異步 flush 模式,在新的 InnoDB flush 機制下以前的問題已經修復。不過 Dimitry 告訴我需要一個更大的 InnoDB 日志文件來獲得更穩定的結果。
下面是 2x4GB vs 2x8GB innodb 日志文件大小的比較:

很顯然,使用更大的日志文件,測試結果更穩定!
結論:
innodb_buffer_pool_instances 參數顯著的影響測試結果,特別是非常高的 I/O 負載時。
在 MySQL 5.6 ,最終是可以獲得非常穩定的吞吐,但自適應的 flush 機制仍需較大的日志文件。
MySQL 配置如下:
04 |
innodb_file_per_table = true |
05 |
innodb_data_file_path = ibdata1:100M:autoextend |
06 |
innodb_flush_method = O_DIRECT |
07 |
innodb_log_buffer_size = 256M |
09 |
innodb_flush_log_at_trx_commit = 1 |
10 |
innodb_buffer_pool_size = 125G |
11 |
innodb_buffer_pool_instances=8 |
13 |
innodb_log_file_size = 4G |
14 |
innodb_log_files_in_group = 2 |
16 |
innodb_read_io_threads = 16 |
17 |
innodb_write_io_threads = 16 |
18 |
innodb_io_capacity = 20000 |
19 |
innodb_io_capacity_max = 40000 |
25 |
max_connections = 2000 |
26 |
max_prepared_stmt_count=500000 |
27 |
max_connect_errors = 10 |
28 |
table_open_cache = 2048 |
29 |
max_allowed_packet = 16M |
30 |
binlog_cache_size = 16M |
31 |
max_heap_table_size = 64M |
34 |
thread_cache_size = 1000 |
45 |
read_rnd_buffer_size = 4M |
46 |
bulk_insert_buffer_size = 8M |
47 |
myisam_sort_buffer_size = 8M |
48 |
myisam_max_sort_file_size = 10G |
49 |
myisam_repair_threads = 1 |
轉自:http://www.phpchina.com/archives/view-41968-1.html