Uber-Update-Stats: keep stats up-to-date on large tables

03 January 2011

Auto_Update_stats: Not helpful on large tables

Experienced DBAs know that auto-update-stats is not sufficient to keep statistics updated on large tables. This is doubly true for tables where the DML activity is near one end of the clustered key.

There are some common workarounds:

  • Have all DML operations run an update statistics command after they’re done. However, this can require a lot of code changes.
  • Update all statistics on a schedule, using sp_updatestats. However, this can lead to bad blocking, and often can’t be run often enough.
  • Swearing at your systems. However,  too much will get you put in management.

Not content with any of these, I’ve gone ahead and built my own script:

Uber-Update-Stats

Script: Uber-Update-Stats.sql Unit Test: Uber-Update-Stats-Test.sql

Now I can run this script on a near-constant schedule, and it will update each table as soon as a configurable percentage of rows or number of rows have changed.  You can also configure the sampling percentages. I think it is The New Hotness very useful.

Coming up next: A version that compensates for filtered indexes & filtered statistics.

Permalink

Introductions, a.k.a. sp_who_are_you

28 December 2010

I’d like to introduce a man with a lot of charm, talent, and wit. Unfortunately, he couldn’t be here tonight, so instead . . .” – Melvin Helitzer

My name is Dev Nambi, and this my blog on databases, data architecture, and solving interesting puzzles. In my short career I’ve grown from watching databases and paging DBAs, to being a pseudo-DBA myself, to writing the code that makes the databases tick. My next step is to become more involved in the SQL Server community. Depending on who I’m talking to, that’s either cause for celebration or agreat reason to join a 12-step program.

"Insanity - a perfectly rational adjustment to an insane world."

Why?

I love working with the people. My mentors and peers are some of the best in the business, including the architect David Denz, the famous Kendra Little (@Kendra_Little) of LittleKendra.com, the infamous Crys Manson (@crysmanson) of Crysmanson.com, the classy Yanni Robel (@YanniRobel) of SqlScribbles.com, the hilarious Argenis Fernandez (@afernandez) of sqlps.com, and the Über-nerd Andrew Rogers (@sqlshaman).

Why else?

I love thorny problems, and big challenges. My biggest joys are performance tuning and figuring out database architecture for large systems.

Permalink