当前位置：网站首页>Database performance analysis and optimization (internal training materials of Aite future team)

Database performance analysis and optimization (internal training materials of Aite future team)

2022-07-28 17:41:00 【itest_ two thousand and sixteen】

Memory management ：MSMM、ASMM(sga_target)、AMM(memory_target) Small memory has the problem of small memory , Large memory has the trouble of large memory ！ORA-04031 Buffer cache and sharedpool size Of begin/end Values in ASMM、AMM and 11gR2MSMM But it will move

Parses Analytic number , Including soft parsing + Hard parsing , What we want is Parse once Run anywhere ！HardParses The root of all evil ．Cursorpin s on X, librarycache: mutex X , latch: row cache objects /sharedpool……………... Hard parsing should preferably be less than per second 20 Time

log file sync： Generally, such waiting time is due to LGWR Progress talk redo log buffer write in redo log Occur in the .

commit Are there too many times

Some components require high response time , During the test, we should set the slow query threshold , Set as small as possible .

yes , we have SQL It may be used for statistics , There is no real-time requirement , At this time, during the test , The slow query threshold can be set higher .

Statspack- Generate snapshot

Suggest : Calculate the usage rate ：10% < Number of connections usage < 85% .

10046 trace To keep track of Oracle

1、 Index is used to improve query SQL The performance of the , But in some cases, queries cannot use indexes , It's useless to build it

2、 Next, let's take a look at our common query types .

3、 Different types of queries , What should we pay attention to when building an index

1、 Different storage engines support different indexing methods

2、 Here we use InnoDB engine , Supported by Btree Take index as an example

Here, we not only query conditions firstname To index , The sorted fields also need to be indexed

1、 Use group by When grouping , It needs to be indexed .

2、 Use distinct When making a uniqueness query ,firstname Although fields are not used as query criteria , Need to add index .

Left connection ： Based on the left table （ The left table takes all the data ）, Go to the right table and match the data by conditions . The search criteria are activated in the right table , So the index is built in the right table .