使用event api诊断数据导入imp性能

王朝other·作者佚名 2008-05-31

宽屏版字体: 小|中|大|超大

SQL> select event,TOTAL_WAITS, TIME_WAITED,AVERAGE_WAIT from v$session_event where sid=18 order by TIME_WAITED desc;

EVENT TOTAL_WAITS

---------------------------------------------------------------- -----------

TIME_WAITED AVERAGE_WAIT

----------- ------------

db file sequential read 47724914

1252067 .026235081

free buffer waits 215054

527065 2.45084955

log file switch completion 85632

397213 4.63860473

EVENT TOTAL_WAITS

---------------------------------------------------------------- -----------

TIME_WAITED AVERAGE_WAIT

----------- ------------

log file sync 388381

213054 .548569575

SQL*Net message from client 9706

87956 9.06202349

latch free 43258

74329 1.71827176

EVENT TOTAL_WAITS

---------------------------------------------------------------- -----------

TIME_WAITED AVERAGE_WAIT

----------- ------------

log file switch (checkpoint incomplete) 1110

27605 24.8693694

SQL*Net more data from client 5254594

22194 .004223733

enqueue 4787

14258 2.97848339

EVENT TOTAL_WAITS

---------------------------------------------------------------- -----------

TIME_WAITED AVERAGE_WAIT

----------- ------------

undo segment extension 9822757

11435 .001164133

write complete waits 928

2936 3.1637931

buffer busy waits 133365

1382 .010362539

EVENT TOTAL_WAITS

---------------------------------------------------------------- -----------

TIME_WAITED AVERAGE_WAIT

----------- ------------

SQL*Net message to client 9706

4 .000412116

file open 22

1 .045454545

发现db file sequential read 居高，通常在单块读发生该事件，用于索引读取；察看正在导入数据的表，果然索引俱全；导入数据的时候要维护索引，对每个导入的数据都要找到对应的索引叶结点插入新索引enry. 删除索引后，该等待事件降低。

通常假如发生较高的log file sync 事件，表示导入进程提交过于频繁。增加buffer参数可以减少commit次数，减少该等待事件