SQL SERVER – Say No To Database Engine Tuning Advisor

I review Database Engine Tuning Advisor recommendations against the workload before applying them. That was the point behind my original opinion.

Overlapping braces are reviewed beside selected fittings and preserved original pieces.

In one consulting case, a table accumulated many suggested structures. We removed more than 100 _DTA indexes and statistics after reviewing their impact. Performance improved in that case. The prefix alone didn’t establish safe removal.

DTA considers indexes, indexed views and partitioning, depending on options and release. I previously described its scope too narrowly. Representative workloads and tested recommendations can make it useful.

My preference for a small index set, often around five, is a heuristic. It is not a SQL Server limit. Balance read benefit with writes, storage, maintenance and constraints. Usage DMVs can reset or miss occasional important work.

Review overlapping indexes, plans and application evidence on a restored copy. Measure the change and preserve a rollback definition. Treat missing-index and unused-index reports as inputs with collection limits.

Reference: DTA workloads and design options.

Related reading

A tuning recommendation is not a deployment instruction, it is a proposal to evaluate against representative work.

Published by Pinal Dave on SQLAuthority. More of my work at pinaldave.com.


Discover more from SQL Authority with Pinal Dave

Subscribe to get the latest posts sent to your email.

SQL Index, SQL Server, SQL Statistics
Previous Post
WRITELOG Waits: Finding What Slows Transaction Log Writes
Next Post
Update Statistics With FULLSCAN and MAXDOP in SQL Server

Related Posts

6 Comments. Leave new

  • I wouldn’t consider myself a follower of yours but so many of my questions end up answered on pages of your blog. Thank you for writing these and please continue.

    Reply
  • Mohammed Haneef
    February 7, 2019 8:48 am

    Thanks for insight as always …. Huge Fan of your blog

    Reply
  • why DTA suggesting extra statistics.
    Generally if we create any index automatically statistics should be created for that particular index. But why again DTA suggesting exta statistics

    Reply
  • If not the DTA, then what do you recommend? The suggestion to “adopt DMVs which gives metadata of the data” is quite vague. Do you have a blog post about how one goes about optimizing in this manner?

    Reply

Leave a Reply

Your email address will not be published. Required fields are marked *

Fill out this field
Fill out this field
Please enter a valid email address.