what performance metrics to show before and after optimization/tuning - to show improvement

Question

Post reply

what performance metrics to show before and after optimization/tuning - to show improvement

VoldemarG

Hall of Fame

Points: 3883
More actions
January 2, 2014 at 1:58 pm

#280858

question: If I am optimizing performance of a few dozen of SPs that i identified via Profiler Trace as taking most time while being the ones most often executed,
and did certain things to improve performance for particularly those SPs. Created some indexes, modified some TSQL inside those SPS, etc.
Now, after I roll out those changes to Production, i will need to show that my improvements/optimization actually has done something, and the performance/metrics are now improved.
Is the best way to prove that changes brought about positive results is to also run Profiler Trace and note that the same SPs ave(duration) and READS have become smaller values? Or is there something else that is useful to
show that modifications (mainly a few dozen of new indexes) have worked and performance is improved? I am thinking what would be most useful to include into my report to management that will show improvement after this tuning/optimization.
Thanks.
Likes to play Chess

Viewing 5 posts - 1 through 4 (of 4 total)

You must be logged in to reply to this topic. Login to reply

Ravid_ds Ten Centuries Points: 1149 More actions · Answer 1

You can test by mentioning SET STATISTICS ON/OFF while executing the SP.

Jeff Moden SSC Guru Points: 1003853 More actions · Answer 2

Ravid_ds (1/3/2014)
You can test by mentioning SET STATISTICS ON/OFF while executing the SP.

Oh be careful now. SET STATISTICS is a known liar when it comes to certain forms of code. Please see the following article for proof of that.

http://www.sqlservercentral.com/articles/T-SQL/91724/

--Jeff Moden

RBAR is pronounced "ree-bar" and is a "Modenism" for Row-By-Agonizing-Row.
First step towards the paradigm shift of writing Set Based code:
________Stop thinking about what you want to do to a ROW... think, instead, of what you want to do to a COLUMN.

Change is inevitable... Change for the better is not.

Helpful Links:
How to post code problems
How to Post Performance Problems
Create a Tally Function (fnTally)

Jeff Moden SSC Guru Points: 1003853 More actions · Answer 3

VoldemarG (1/2/2014)
question: If I am optimizing performance of a few dozen of SPs that i identified via Profiler Trace as taking most time while being the ones most often executed,
and did certain things to improve performance for particularly those SPs. Created some indexes, modified some TSQL inside those SPS, etc.
Now, after I roll out those changes to Production, i will need to show that my improvements/optimization actually has done something, and the performance/metrics are now improved.
Is the best way to prove that changes brought about positive results is to also run Profiler Trace and note that the same SPs ave(duration) and READS have become smaller values? Or is there something else that is useful to
show that modifications (mainly a few dozen of new indexes) have worked and performance is improved? I am thinking what would be most useful to include into my report to management that will show improvement after this tuning/optimization.
Thanks.

You need to show both because the indexes and other changes you make may have an impact on other things. Here's a URL to help you establish a baseline and then check and see what happens after the changes.

http://www.sqlservercentral.com/articles/baselines/94656/

--Jeff Moden

RBAR is pronounced "ree-bar" and is a "Modenism" for Row-By-Agonizing-Row.
First step towards the paradigm shift of writing Set Based code:
________Stop thinking about what you want to do to a ROW... think, instead, of what you want to do to a COLUMN.

Change is inevitable... Change for the better is not.

Helpful Links:
How to post code problems
How to Post Performance Problems
Create a Tally Function (fnTally)

VoldemarG Hall of Fame Points: 3883 More actions · Answer 4

Thank you Jeff.

Great article on base-lining that you have referred to. http://www.sqlservercentral.com/articles/baselines/94656/ [/url]

1) So ideally base-lining would be an ongoing (daily) process with a job setup

that will write a snapshot of config + performance data to base-line tables?

2) and how about the query/SP performance data? The one that comes from DMVs. It does not appear too useful to store a snapshot of such info as the one from sys.dm_exec_query_stats or sys.dm_exec_procedure_stats on a daily basis because the metrics returned by this view get reset when Server restarts and SP is recompiled, etc. ? In other words, my main question is is what's the best way (in addition to Profiler Trace data) to also include an ongoing query performance info into the kind of recurring base-line data collection like the one referred to in the above article?

Likes to play Chess