Window Aggregate operator in batch mode in SQL Server 2019

So this came as a surprise, when working on calculating simple statistics on my dataset, in particular min, max and median. First two are trivial. The last one was the one, that caught my attention.

While finding the fastest way on calculating the median (statistic: median) for given dataset, I have stumbled upon an interesting thing.  While WINDOW function was performing super slow and calling R or Python using sp_execute_xternal_script outperform window function as well, it raised couple of questions.

But first, I created a sample table and populate it sample rows:

DROP TABLE IF EXISTS  t1;
GO

CREATE TABLE t1
(id INT IDENTITY(1,1) NOT NULL
,c1 INT
,c2 SMALLINT
,t VARCHAR(10) 
)

SET NOCOUNT ON;
INSERT INTO t1 (c1,c2,t)
SELECT 
	x.* FROM
(
	SELECT 
	ABS(CAST(NEWID() AS BINARY(6)) %1000) AS c1
	,ABS(CAST(NEWID() AS BINARY(6)) %1000) AS c2
	,'text' AS t
) AS x
	CROSS JOIN (SELECT number FROM master..spt_values) AS n
	CROSS JOIN (SELECT number FROM master..spt_values) AS n2
GO 2

 

Query generated – in my case – little over 13 million records, just enough to test the performance.

So starting with calculating Median, but sorting first half and second half of rows respectively, the calculation time was surprisingly long:

-- Itzik Solution
SELECT (
(SELECT MAX(c1) FROM
  (SELECT TOP 50 PERCENT c1 FROM t1 ORDER BY c1) AS BottomHalf)
+
(SELECT MIN(c1) FROM
  (SELECT TOP 50 PERCENT c1 FROM t1 ORDER BY c1 DESC) AS TopHalf)
) / 2 AS Median

Before and after each run, I cleaned the stored execution plan. The execution on 13 million rows took – on my laptop – around 45 seconds.

Next query, for median calculation was a window function query.

SELECT DISTINCT
PERCENTILE_CONT(0.5) WITHIN GROUP (ORDER BY c1) 
       OVER (PARTITION BY (SELECT 1)) AS MedianCont
FROM t1

To my surprise, the performance was even worse, and at this time, I have to say, I was running this on SQL Server 2017 with CU7. But luckily, I had a SQL Server 2019 CTP 2.0 also installed and here, with no further optimization the query ran little over 1 second.

So the difference between the versions was enormous. I could replicate the same results by switching the database compatibility level from 140 to 150, respectively.

ALTER DATABASE SQLRPY 
SET COMPATIBILITY_LEVEL = 140; 
GO
SELECT DISTINCT
PERCENTILE_CONT(0.5) WITHIN GROUP (ORDER BY c1) 
    OVER (PARTITION BY (SELECT 1)) AS MedianCont140
FROM t1

ALTER DATABASE SQLRPY 
SET COMPATIBILITY_LEVEL = 150; 
GO

SELECT DISTINCT
PERCENTILE_CONT(0.5) WITHIN GROUP (ORDER BY c1) 
    OVER (PARTITION BY (SELECT 1)) AS MedianCont150
FROM t1

The answer was found in execution plan. When running window function under 140 compatibility level, execution plan decides to create nested loop two times, for both groups of upper and lower 50% of the dataset.comp140_1

This plan is is somehow similar to understanding of 50% of upper and lower dataset but with only one nested loop:

itzik1

Difference is that when running the window function calculation of median on SQL Server version 2017, the query optimizer decides to take row execution mode for built-in window function with WITHIN GROUP.

comp140_2

This was, as far as I knew, not an issue since SQL Server 2016, where batch mode operator for window aggregation was already used.

When switching to compatibility level 150 and running the same window function, the execution plan is, as expected:

comp150_1

And window aggregate uses batch mode:

comp150_2

When calculating Median using R:

sp_Execute_External_Script
   @language = N'R'
  ,@script = N'd <- InputDataSet
               OutputDataSet <- data.frame(median(d$c1))'
  ,@input_data_1 = N'select c1 from t1'
WITH RESULT SETS (( Median_R VARCHAR(100) ));
GO

or Python:

sp_Execute_External_Script
  @language = N'Python'
 ,@script = N'
import pandas as pd
dd = pd.DataFrame(data=InputDataSet)
os2 = dd.median()[0]
OutputDataSet = pd.DataFrame({''a'':os2}, index=[0])'
 ,@input_data_1 = N'select c1 from t1'
WITH RESULT SETS (( MEdian_Python VARCHAR(100) ));
GO

both are executing and returning the results in about 5 seconds. So no bigger difference between R and Python when handling 13 million rows for calculating simple statistics.

To wrap up, If you find yourself in situation, where you need to calculate – as in my case – Median or any statistics, using window function within group, R or Python would be the fastest solutions, following T-SQL. Unless, you have the ability to use SQL Server 2019, T-SQL is your best choice.

Code and the plans, used in this blog post are available, as always at Github.

Tagged with: , , , , , , , , ,
Posted in Uncategorized
2 comments on “Window Aggregate operator in batch mode in SQL Server 2019
  1. […] leave a comment for the author, please follow the link and comment on their blog: R – TomazTsql. R-bloggers.com offers daily e-mail updates about R news and tutorials on topics such as: Data […]

    Like

  2. […] Tomaz Kastrun shows that batch aggregation mode on window functions allow PERCENTILE_CONT finally to become useful: […]

    Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

Categories
Follow TomazTsql on WordPress.com
Programs I Use
Programs I Use
Programs I Use
Rdeči Noski – Charity

Rdeči noski

100% of donations made here go to charity, no deductions, no fees. For CLOWNDOCTORS - encouraging more joy and happiness to children staying in hospitals (http://www.rednoses.eu/red-noses-organisations/slovenia/)

€2.00

Top SQL Server Bloggers 2018
TomazTsql

Tomaz doing BI and DEV with SQL Server and R, Python and beyond

Discover

A daily selection of the best content published on WordPress, collected for you by humans who love to read.

Revolutions

Tomaz doing BI and DEV with SQL Server and R, Python and beyond

tenbulls.co.uk

attaining enlightenment with sql server, .net, biztalk, windows and linux

SQL DBA with A Beard

He's a SQL DBA and he has a beard

Reeves Smith's SQL & BI Blog

A blog about SQL Server and the Microsoft Business Intelligence stack with some random Non-Microsoft tools thrown in for good measure.

SQL Server

for Application Developers

Business Analytics 3.0

Data Driven Business Models

SQL Database Engine Blog

Tomaz doing BI and DEV with SQL Server and R, Python and beyond

Search Msdn

Tomaz doing BI and DEV with SQL Server and R, Python and beyond

R-bloggers

Tomaz doing BI and DEV with SQL Server and R, Python and beyond

Ms SQL Girl

Julie Koesmarno's Journey In Data, BI and SQL World

R-bloggers

R news and tutorials contributed by hundreds of R bloggers

Data Until I Die!

Data for Life :)

Paul Turley's SQL Server BI Blog

sharing my experiences with the Microsoft data platform, SQL Server BI, Data Modeling, SSAS Design, Power Pivot, Power BI, SSRS Advanced Design, Power BI, Dashboards & Visualization since 2009

Grant Fritchey

Intimidating Databases and Code

Madhivanan's SQL blog

A modern business theme

Alessandro Alpi's Blog

SQL Server, Azure and DLM in a nutshell :D

Paul te Braak

Business Intelligence Blog

Sql Server Insane Asylum (A Blog by Pat Wright)

Information about SQL Server from the Asylum.

Gareth's Blog

A blog about Life, SQL & Everything ...

SQLPam's Blog

Life changes fast and this is where I occasionally take time to ponder what I have learned and experienced. A lot of focus will be on SQL and the SQL community – but life varies.

%d bloggers like this: