Interesting, I hadn't previously heard that.
Makes sense though, indexes work that way by default. And you can't change that behavior. Using a combination of PCTFREE and PCTUSED that adds up to 100 can lead to a lot of update activity on the free list. Or at least I seem to recall that, I haven't actually experienced it. At least I dont' remember experiencing it. Everyone was a newbie at some point. :) Jared On Saturday 20 April 2002 09:33, Don Granaman wrote: > For the radical approach - as espoused by Dave Ensor - PCTUSED=0 !!! (If > this doesn't generate some serious discussion, I'll be amazed!) > > This works quite well for tables where records are purged some set length > of time after insert. The net result is that blocks are filled, then left > alone until purged when they (the vast majority at least) are completely > emptied and put back on the free list. Blocks tend to be fully populated > and freelist activity is minimized. Be aware that in some older versions > of Oracle, there is a bug that requires PCTUSED = 1 (blocks only move back > to the free list if used space < PCTUSED instead of <= PCTUSED). I do not > know offhand what version fixed this. > > Dave Ensor's suggestion was to use this (almost) everywhere, but I'm not > really comfortable with it as a blanket policy. It seems that tables with > deletes that are "chronologically random" would end up with a LOT of very > sparsely populated blocks. > > The approach recommended below is to set it rather high to accomplish the > block density goal, but at the potential expense of high freelist activity. > Which approach is more appropriate depends on the nature of one's > insert/delete criteria. > > Also, PCTFREE can be set extremely low (1,5,?) for some kinds of objects - > static tables, indexes with strictly inorder data (CREATE_DATE, SEQUENCE#, > etc.). > > I agree with always setting PCTINCREASE =0 > > Don Granaman > [OraSaurus - Honk if you remember UFI !] > > ----- Original Message ----- > To: "Multiple recipients of list ORACLE-L" <[EMAIL PROTECTED]> > Sent: Thursday, April 18, 2002 8:08 AM > > > Seema, > > > > PCTINCREASE - always 0. > > > > PCTFREE - for lookup (code) tables = 10 for others, I use 20. My theory > > is > > > that lookup tables have little or no updates, so this value should be > > small. > > > If you have a high-update tables (columns that are updated a 8lot* after > > the > > > record was created, you may want a higher value here. > > > > PCTUSED - for lookup (code) tables = 90, for others, I use 80. Again, > > for lookup tables with little or no updates, I fill the data blocks as > > high as possible. For other tables, I currently use 80 as a starting > > point. > > > > You will get many opinions here - read the docs to determine what you > > think > > > is best for you. > > > > Hope this helps. > > > > Tom Mercadante > > Oracle Certified Professional -- Please see the official ORACLE-L FAQ: http://www.orafaq.com -- Author: Jared Still INET: [EMAIL PROTECTED] Fat City Network Services -- (858) 538-5051 FAX: (858) 538-5051 San Diego, California -- Public Internet access / Mailing Lists -------------------------------------------------------------------- To REMOVE yourself from this mailing list, send an E-Mail message to: [EMAIL PROTECTED] (note EXACT spelling of 'ListGuru') and in the message BODY, include a line containing: UNSUB ORACLE-L (or the name of mailing list you want to be removed from). You may also send the HELP command for other information (like subscribing).
