Thanks, Karl, for your very prompt response. I apologize for my less-than-immediate reply.
After a discussion with my developer, I realized that we ARE modifying existing records rather than deleting and then re-inserting (most of) them. The number of deleted records (and orphaned btree keys) is therefore insignificant in this 24M-record table. That renders moot my rationale for the modify... to merge operation. I appreciate your time and efforts to address my question, and will consider the modify... to reconstruct action along with the other options to maintain and optimize our storage and access objects. Best regards, Dave Marvin Columbia Basin PIT Tag Information System Pacific States Marine Fisheries Commission dave.mar...@ptagis.org -----Original Message----- From: users-boun...@lists.ingres.com [mailto:users-boun...@lists.ingres.com] On Behalf Of Karl & Betty Schendel Sent: Thursday, August 06, 2009 10:53 AM To: users@lists.ingres.com Subject: Re: [Users] Modify... to merge option to shrink a btree index I'd advise against a modify to merge. If you want to try something, just modify the table to reconstruct. Modify to merge is very log- intensive, and outside of certain specific cases I can't recommend it. It's hard to say for sure when a btree needs modified; it's usually easier to just do it rather than trying to analyze it. (you can get into tools like "modify to table_debug" but by the time you have interpreted the output, your modify to reconstruct would be finished...) Karl On Aug 6, 2009, at 1:36 PM, Dave Marvin wrote: > Hello: > > > > A table of 24M records is keyed on a btree index of unique > values. Records are generally appended to this table. Records > are occasionally replaced (rather than modified) by deleting and > then appending similar or even identical data. Rarely are records > permanently deleted from the table. > > > > Will the btree index on this table benefit from a merge? How can I > tell? Are there graphical or tabular metrics that I can review to > show me the number of inactive (deleted) index values relative to > the entire index structure? Knowing how much the index will > shrink, and how long it takes to regenerate the btree index from > scratch, can I reliably guestimate how long the modify... to merge > will take? > > > > Cheers! > > Dave > > > > Dave Marvin > > Columbia Basin PIT Tag Information System > > Pacific States Marine Fisheries Commission > > dave.mar...@ptagis.org > > > > _______________________________________________ > Users mailing list > Users@lists.ingres.com > http://lists.ingres.com/mailman/listinfo/users _______________________________________________ Users mailing list Users@lists.ingres.com http://lists.ingres.com/mailman/listinfo/users _______________________________________________ Users mailing list Users@lists.ingres.com http://lists.ingres.com/mailman/listinfo/users
