Re: Importing huge data

"Howard J. Rogers" <howardjr2000_at_yahoo.com.au> wrote in message news:3f550339$0$14559$afc38c87_at_news.optusnet.com.au...
> Bear in mind that these two methods will succesfully switch off the
> production of redo for your particular set of inserts. But *nothing*
> switches off the production of rollback. So that's still an overhead you
> will have to live with, and the only thing you can do there to speed
things
> up would be to make sure that your rollback segments are big enough to
start
> with that they never need to grow during the loading process itself, and
> that they are housed within a rollback tablespace which is not using the
> same disk resources as the table you are inserting into (so that you avoid
> I/O contention issues).

Howard,

doesn't a direct path load or INSERT /*+ APPEND */ usually generate less rollback than a convention load / insert ?

My recollection is that a direct path load alway inserts its data at the end of the table (even if there is free space elsewhere) and following the load either a commit or rollback must be issued before any other data in the table can be added or modified. Because of this the rollback only needs a pointer to the range of rows that have been added, and the insert of a large volume of data could generate very little rollback.

Of course I could be completely wrong about this.

Paul Dixon Received on Wed Sep 03 2003 - 05:15:38 CDT