Improve performance of create statements on a highly fragmented disk [message #483760] |
Wed, 24 November 2010 07:39  |
steffeli
Messages: 112 Registered: July 2006
|
Senior Member |
|
|
Hi,
On Oracle 10g, I create, delete and drop a lot of tables. Therefore, the disk is highly fragmented.
The execution of a very simple create statement takes more than a minute. If I execute the same statement but first truncate the table and insert the data, it takes less than a second!
I think this has to do with the high fragmentation of the disk. Obviously, I can defragment the disk, but I will always have a high fragmentation since I use a lot of create, delete and drops.
Any idea how I can improve the performance of create statements on highly fragmented disks?
Thanks, Stefan
[Updated on: Wed, 24 November 2010 07:41] Report message to a moderator
|
|
|
|
|
|
|
|
|
Re: Improve performance of create statements on a highly fragmented disk [message #483796 is a reply to message #483777] |
Wed, 24 November 2010 09:33  |
Roachcoach
Messages: 1576 Registered: May 2010 Location: UK
|
Senior Member |
|
|
Actually, unless I'm mistaken, disk performance shouldnt (seriously) impact create table as/insert - Doesnt DBW write independently of the code executing? Unless you're experiencing poor read speed in the create, but from your last reply to me that would suggest its not the case.
I could be wrong, its been a while since I looked at this in depth, but I think that the CTAS and insert should only be disc write bottlenecked by the log buffer disc (and if you're not commiting the insert that might appear to remove that bottleneck).
Doubtless one of the other folks here can confirm/deny.
[Updated on: Wed, 24 November 2010 09:34] Report message to a moderator
|
|
|