Re: Interesting problem
From: mhoys <matthias.hoys_at_gmail.com>
Date: Thu, 26 May 2011 00:26:52 -0700 (PDT)
Message-ID: <f49e395e-fb50-4262-b513-910fbe8a08c2_at_x10g2000yqj.googlegroups.com>
> As for the cursors and bulk operations, that was the result of my
> benchmark. This was simply faster than the explicit delete. There can be
> quite a few rows to delete, so it must be done in a piecewise fashion.
> Deleting batches of rows using bulk operations turned to be faster than
> using normal "delete" command, but I am not sure why. I do remember
> testing it, and it was consistently faster option.
>
Date: Thu, 26 May 2011 00:26:52 -0700 (PDT)
Message-ID: <f49e395e-fb50-4262-b513-910fbe8a08c2_at_x10g2000yqj.googlegroups.com>
> As for the cursors and bulk operations, that was the result of my
> benchmark. This was simply faster than the explicit delete. There can be
> quite a few rows to delete, so it must be done in a piecewise fashion.
> Deleting batches of rows using bulk operations turned to be faster than
> using normal "delete" command, but I am not sure why. I do remember
> testing it, and it was consistently faster option.
>
That's strange. My initial thought was also "couldn't this be more performant with 2 simple delete statements?". I did a quick test with a table with 1 million records (agreed, nothing big) and a single delete statement was 50% faster than a cursor with bulk operation (database is 10.2.0.5). You wrote "it must be done in a piecewise fashion". What is the reason for this?
Matthias Received on Thu May 26 2011 - 02:26:52 CDT