|
On OLTP, it's a bad idea to always increase shared pool when its free memory (from v$sgastat) is low because there may be too many SQLs not using bind variables. But for data warehouses, its free memory becomes a more meaningful indicator.
Any sorting uses temp tablespace, including order by, group by, sort merge join. You don't want to leave its datafile autoextensible because you don't want a file larger than 2G (somewhere there's still such limit).
Other comments. If I remember right, UGA is in PGA for dedicated server and in SGA for MTS. Too large log buffer may cause log file sync wait. Usually larger than 1M is a waste of memory and harmful in pre-8i (according to Steve Adams). "Larger values reduce log file I/O" should be changed to "Larger values reduce the number of log file I/Os but may increase the time of every I/O".
Yong Huang
最初由 ligengocp 发布
[B]
所以,我的意思是,虽然你的系统中没有像我这般傻的SQL语句,但如果有查询很多,且查询时间很长(6,7个小时以上)时,你的shared_pool大一些比较好。
...
此外,temp表空间你最好不要设置成自动增长,我曾经有一次TEMP表空间文件增长到5G,我还不知道。
TEMP表空间好像是查询时有ORDER BY时会使用,不知是否正确,还望指正。 [/B] |
|