![]() | |
#1
| |||
| |||
|
#2
| |||
| |||
|
|
Hi, We want to pin a couple of tables, hoping to increase some performance. *From the docs, a pinned table means less reads from disk, etc. If the table gets updated or inserted into, is there any advantage to this? *If yes, how does the committed transactions get saved? |
#3
| |||
| |||
|
|
On Jul 20, 11:57*am, The Magnet <a... (AT) unsu (DOT) com> wrote: Hi, We want to pin a couple of tables, hoping to increase some performance. *From the docs, a pinned table means less reads from disk, etc. If the table gets updated or inserted into, is there any advantage to this? *If yes, how does the committed transactions get saved? Why do you not consider using the KEEP pool? http://oratips-ddf.blogspot.com/2008...ts-keeper.html David Fitzjarrell |
#4
| |||
| |||
|
|
Hi, We want to pin a couple of tables, hoping to increase some performance. *From the docs, a pinned table means less reads from disk, etc. If the table gets updated or inserted into, is there any advantage to this? *If yes, how does the committed transactions get saved? |
#5
| |||
| |||
|
|
On Jul 20, 9:57*am, The Magnet <a... (AT) unsu (DOT) com> wrote: Hi, We want to pin a couple of tables, hoping to increase some performance. *From the docs, a pinned table means less reads from disk, etc. If the table gets updated or inserted into, is there any advantage to this? *If yes, how does the committed transactions get saved? If you just let Oracle do its thing, blocks that get used a lot will stay in memory, since Oracle uses an LRU algorithm. *Also, small tables use a different algorithm for full scanning ("The definition of a small table is the maximum of 2% of the buffer cache and 20, whichever is bigger.") *I used to have noticeable results on certain objects with a recycle pool, but nowadays don't seem to need to bother. *Be careful about catching obsessive tuning disorder. *Take any rule of thumb that uses percentages with a very large dose of salt. *Remember that most performance problems come from the app code. *"Hoping" is not a particularly good tuning methodology. *You want to use a methodology that tells you how to find what is wrong and where to put your effort. See metalink Note: 135223.1 and note the auto-tuning part. *And keep a lot of salt handy. See commit transactions in the concepts manual for the basic idea on how that works. Do you have an actual problem to solve? *You need to state it exactly. *http://dbaoracle.net/readme-cdos.htm jg -- @home.com is bogus.http://www3.signonsandiego.com/stori...judges23536-fe... |
![]() |
| Thread Tools | |
| Display Modes | |
| |