Re: Efficiency of sqlldr

Nan wrote:
> The primary difference between options 1 and 2 is the way in which
> data "grows" in the tables. Option 1 one table is populated fully at a
> time. In option 2 each table grows by 1000 rows at a time.
> Does this cause the physical location of the data on the disks to be
> different in options 1 and 2 ? If yes, then select queries with joins
> on the tables will be more efficient if data was loaded using option
> #2 than option #1.
> Any opinions are appreciated.

It depends on a lot of factors, some of them probably out of your control. e.g.

• RAID or conventional disk?
• How does your file system work?
• What are the storage parameters for the tablespace and datafiles?
• Are the datafiles fragmented?
• What are the storage parameters for the table?
• Do you use indices? Again, what are the storage parameters?
• Do you sort the data before loading?
• ...

I'd go with option #1, just for the reason that it's less effort. Let Oracle do the handling instead of trying to secondguess what it does.

Stephan Received on Thu May 27 2004 - 08:33:32 CDT