IDS Forum

Re: update stats high question

Posted By: DL Redden
Date: Wednesday, 21 June 2006, at 11:04 a.m.

In Response To: Re: update stats high question (ART KAGEL, BLOOMBERG/ 731 LEXIN)

What is the benefit of running an upstats low on each complete index key if the same column is used in multiple indexes. It seems to me that upstats low would then be generated on that column multiple times resulting in unnecessary work.

----- Original Message ----
From: "ART KAGEL, ...." <kagel@bloomberg.net>
To: ids@iiug.org
Sent: Wednesday, June 21, 2006 8:03:11 AM
Subject: Re: update stats high question [7013]

The algorithm recommended in the Performance Guide and in John Miller III's
paper, and implemented in dostats, is designed to minimize the runtimes of the
update stats runs while providing sufficiently detailed stats for the
optimizer.
If you are running HIGH on the entire table that is ALMOST enough, and will,
for non-key leading columns, provide better stats than dostats, but will tend
to run a bit longer. The only thing I think you will be missing are the LOW
calculations for each index key. Perhaps John Miller can comment on that. So,
based on my assertion, I would add to the HIGH's a LOW on each complete index
key.

Art S. Kagel

----- Original Message -----
From: Doug Fossmeyer <ids@iiug.org>
At: 6/20 16:09:28

I have a couple of functional as well as architectural questions for the
update statistics gurus. After performing a migration from IDS 7 to IDS 94fc6
it was recommended in the performance and migration manuals to drop dists and
run update stats as appropriate. It was also suggested we only need to run
update statistics high for the smaller tables.
- One area I did not see addressed is that the high statement will implicitly
perform the drop distributions as part of it's process?
- Does this imply that simply running update statistics high for any table
(regardless of size, width, performance and time) is the same as the more
discrete and elaborate methods employed by dostats and other utilities for the
indexed columns?
- Is the intent of the various update statistics statements to minimize the
impact on the system as well as the elapsed time to complete?
- Is there an issue with the amount of data the high statement generates in
the various system tables and is it actual counter productive with respect to
the optimizer for a small to moderate size system (3-4 databases each at 24
gig and about 700 tables)?

Thanks in advance,
Doug

*******************************************************************************
Forum Note: Use "Reply" to post a response in the discussion forum.

*******************************************************************************
Forum Note: Use "Reply" to post a response in the discussion forum.

Messages In This Thread

• update stats high question
  Doug Fossmeyer -- Tuesday, 20 June 2006, at 4:05 p.m.
  • Re: update stats high question
    ART KAGEL, BLOOMBERG/ 731 LEXIN -- Wednesday, 21 June 2006, at 9:03 a.m.
    • Re: update stats high question
      DL Redden -- Wednesday, 21 June 2006, at 11:04 a.m.
      • Re: update stats high question
        ART KAGEL, BLOOMBERG/ 731 LEXIN -- Wednesday, 21 June 2006, at 11:08 a.m.
        • RE: update stats high question
          Ford, Andrew G -- Wednesday, 21 June 2006, at 11:41 a.m.

IDS Forum is maintained by Administrator with WebBBS 5.12.