Re: Best Practice: pinning objects
- From: ryan_gaffuri@xxxxxxxxxxx
- To: ajay_1ajay@xxxxxxxxx, sacrophyte@xxxxxxxxx, oracle-l <oracle-l@xxxxxxxxxxxxx>
- Date: Mon, 05 Mar 2007 17:59:16 +0000
I have yet to see a true tangible reason to pin anything in memory. If you are
going to use it alot it will be in memory. If you, don't worry about it. I have
also not seen any performance improvements from the use of pinning at all. It
just seems like extra work and extra maintenance for something that does not
improve performance in any noticeable way.
Reducing the small amount of physical IO you will get from a pinned object that
is not staying in memory because it is not used enough is insignificant.
If you are going to go through the effort of coding a pinning scheme test it to
see if it makes any type of difference.
-------------- Original message --------------
From: Ajay <ajay_1ajay@xxxxxxxxx>
What are the valid points against pinning?
What do you mean by managing the trigger?
You can put the objects to pin in a database table and join it with dba_objects
and pin the appropriate ones. You only change the table to add/remove objects
and leave the trigger code alone. Thanks
Charles Schultz <sacrophyte@xxxxxxxxx> wrote:
Good day, list,
My colleagues and I were debating the merits of having a startup trigger pin
objects. I am in favor of this method, but the others bring some valid points
to the table. It basically comes down to managing the trigger and the objects
that are pinned. We experimented with a shutdown trigger in the past, which was
responsible for gathering stats and collecting info about "current' objects,
but that broke rather quickly. So we are thinking of providing a list of
objects from an external source (ie, flat file), which would be consistent and
extremely visible. The alternative to using a trigger would be a cron job that
checks every-so-often.
I tried to find supporting documentation online, but I did not get any useful
hits for "best practice[s]" and "pin object[s]". Anyone want to weigh in on
either side? One way or another, we want various application objects to be
pinned right after the database is started, with minimal amount of hassle and
overhead. From my point of view, the whole purpose of a startup trigger is to
run something when the database starts. Yes, one could use a glorified startup
script, but we want to steer away from such conventions. One key question is
"Is there ever a time when you do not want to automatically pin objects?"
I have also been looking for information about whether or not a startup trigger
would fire for "STARTUP UPGRADE" when patching.
PS - I have an example I am using from M.Gralike, 2003-10-10
--
Charles Schultz
Be a PS3 game guru.
Get your game face on with the latest PS3 news and previews at Yahoo! Games.
Other related posts:
