Surprised that Row_Number() has better performance than IDENTITY

Question

Surprised that Row_Number() has better performance than IDENTITY

souLTower

SSCarpal Tunnel

Points: 4026
More actions
January 17, 2020 at 1:39 pm

#3716298

I was doing some testing to see if it's more efficient to use an identity column or to use ROW_NUMBER() to simulate a unique ID for the result set. My thinking was that ROW_NUMBER() would need to do some type of evaluation of the result set in order to assign the value where IDENTITY would use the seed and simply increment. Using a result set of 4 million records the ROW_NUMBER() approach was a full second faster. This is the opposite of what I thought would happen. Just an academic observation but I thought it was interesting.

Viewing 6 posts - 1 through 6 (of 6 total)

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

Grant Fritchey SSC Guru Points: 399061 More actions · Answer 1

Not that shocking. To use an IDENTITY column, you need to write data somewhere, whereas the ROW_NUMBER() is just a calculation.

"The credit belongs to the man who is actually in the arena, whose face is marred by dust and sweat and blood"
- Theodore Roosevelt

Author of:
SQL Server Execution Plans
SQL Server Query Performance Tuning

souLTower SSCarpal Tunnel Points: 4026 More actions · Answer 2

Actually I wrote data in both instances. An example is below

-- Approach 1

CREATE TABLE FOO (iid int identity, ky int)

INSERT FOO(ky) SELECT ky FROM myTable ORDER BY ky

-- Approach 2

CREATE TABLE FOO (iid int, ky int)

INSERT FOO(iid, ky) SELECT ROW_NUMBER() OVER(ORDER BY ky) as iid, ky FROM myTable ORDER BY ky

Grant Fritchey SSC Guru Points: 399061 More actions · Answer 3

Try the test again, but use a SEQUENCE. You should see the same results. It's because of the processing needed for maintaining an identity value versus the calculation of the ROW_NUMBER.

"The credit belongs to the man who is actually in the arena, whose face is marred by dust and sweat and blood"
- Theodore Roosevelt

Author of:
SQL Server Execution Plans
SQL Server Query Performance Tuning

souLTower SSCarpal Tunnel Points: 4026 More actions · Answer 4

souLTower

SSCarpal Tunnel

Points: 4026

January 17, 2020 at 3:17 pm

#3716334

Thanks!

ST

Jeff Moden SSC Guru Points: 1004601 More actions · Answer 5

I'm thinking that SEQUENCE is going to be about 4 times slower. That's just a guess.

Sounds crazy but this is fascinating to me because I work with some really large numbers of rows in the things I do and milliseconds do matter here. I'm putting together some of my own tests that I'll share but I've got an appointment I have to go to.

--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)

Surprised that Row_Number() has better performance than IDENTITY

Cookies on SQLServerCentral