Re: Dynamic Cursor
From: joel garry <joel-garry_at_home.com>
Date: Thu, 17 Jun 2010 09:48:14 -0700 (PDT)
Message-ID: <23d195e4-199d-46d5-9866-55cb362f8755_at_k17g2000pro.googlegroups.com>
On Jun 17, 6:07 am, The Magnet <a..._at_unsu.com> wrote:
> Yes, it makes sense. I may have to use that approach, though I always
> hated creating temporary tables and such. I figured if I can do it in
> memory, then it is a better thing to do.
Date: Thu, 17 Jun 2010 09:48:14 -0700 (PDT)
Message-ID: <23d195e4-199d-46d5-9866-55cb362f8755_at_k17g2000pro.googlegroups.com>
On Jun 17, 6:07 am, The Magnet <a..._at_unsu.com> wrote:
>
> Yes, it makes sense. I may have to use that approach, though I always
> hated creating temporary tables and such. I figured if I can do it in
> memory, then it is a better thing to do.
That's the cool thing about Oracle. It does it in memory, then writes it out when it feels like it, which may be before or after you commit. It's optimistic about whether you might commit. So you get the best of both worlds (memory and disk) without having to do anything except not work against it (which includes configuring so it doesn't feel more compelled to write to disk with your data volumes). Randolph shows one way to investigate some details: http://oracle-randolf.blogspot.com/2009/06/temporary-tablespace-groups.html
I've seen a lot of stupid code coming from people who thought they could do better than Oracle, forcing things to memory inappropriately, necessitating way more I/O in the end.
jg
-- _at_home.com is bogus. http://www.signonsandiego.com/news/2010/jun/17/internet-sleuths-get-new-way-to-report-stolen-data/Received on Thu Jun 17 2010 - 11:48:14 CDT