Slow Query

Question

Slow Query

ss-457805

SSCertifiable

Points: 5872
More actions
July 6, 2017 at 9:55 am

#324381

Hi Guys
Aprreciate your help with this. I am trying to tune the following query: It is part of a stored proc.
SELECT s.SupplierID, c.Carrier as Supplier FROM (SELECT DISTINCT SupplierID FROM CPT_SUBDC_Analysis UNION ALL SELECT DISTINCT SupplierID FROM CPT.dbo.CPT_SUBDC_Analysis) s
LEFT OUTER JOIN Carrier c ON c.CarrierID = s.SupplierID
GROUP BY s.SupplierID, c.Carrier
ORDER BY c.Carrier
This is currently running quite frequently. It takes on average 2 minutus to run.
Most of the cost is with this - SELECT DISTINCT SupplierID FROM CPT_SUBDC_Analysis UNION ALL SELECT DISTINCT SupplierID FROM CPT.dbo.CPT_SUBDC_Analysis
Is there another way to use this in the query?
Many thanks in advance.
blog: http://sarveshsingh.com Twitter: @sarveshsing

Viewing 7 posts - 1 through 6 (of 6 total)

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

ChrisM@Work SSC Guru Points: 186127 More actions · Answer 1

ss-457805 - Thursday, July 6, 2017 9:55 AM
Hi Guys
Aprreciate your help with this. I am trying to tune the following query: It is part of a stored proc.
SELECT s.SupplierID, c.Carrier as Supplier FROM (SELECT DISTINCT SupplierID FROM CPT_SUBDC_Analysis UNION ALL SELECT DISTINCT SupplierID FROM CPT.dbo.CPT_SUBDC_Analysis) s
LEFT OUTER JOIN Carrier c ON c.CarrierID = s.SupplierID
GROUP BY s.SupplierID, c.Carrier
ORDER BY c.Carrier
This is currently running quite frequently. It takes on average 2 minutus to run.
Most of the cost is with this - SELECT DISTINCT SupplierID FROM CPT_SUBDC_Analysis UNION ALL SELECT DISTINCT SupplierID FROM CPT.dbo.CPT_SUBDC_Analysis
Is there another way to use this in the query?
Many thanks in advance.

You could try sorting the two concatenated sets only once:
SELECT s.SupplierID, c.Carrier as Supplier FROM ( SELECT DISTINCT SupplierID FROM ( SELECT SupplierID FROM CPT_SUBDC_Analysis UNION ALL SELECT SupplierID FROM CPT.dbo.CPT_SUBDC_Analysis ) d ) s LEFT OUTER JOIN Carrier c ON c.CarrierID = s.SupplierID GROUP BY s.SupplierID, c.Carrier ORDER BY c.Carrier

Also, see UNION without the ALL operator. I prefer the version above.

^{“Write the query the simplest way. If through testing it becomes clear that the performance is inadequate, consider alternative query forms.” - Gail Shaw}

For fast, accurate and documented assistance in answering your questions, please read this article.
Understanding and using APPLY, (I) and (II) Paul White
Hidden RBAR: Triangular Joins / The "Numbers" or "Tally" Table: What it is and how it replaces a loop Jeff Moden

Luis Cazares SSC Guru Points: 183706 More actions · Answer 2

Why are you using DISTINCT and UNION ALL? That might be the the reason you need the GROUP BY.
SELECT s.SupplierID, c.Carrier as Supplier FROM (SELECT SupplierID FROM CPT_SUBDC_Analysis UNION SELECT SupplierID FROM CPT.dbo.CPT_SUBDC_Analysis) s LEFT OUTER JOIN Carrier c ON c.CarrierID = s.SupplierID ORDER BY c.Carrier

Anyway, to get better help, post DDL and execution plans as shown in the following article: How to Post Performance Problems - SQLServerCentral

Luis C.
General Disclaimer:
Are you seriously taking the advice and code from someone from the internet without testing it? Do you at least understand it? Or can it easily kill your server?

How to post data/code on a forum to get the best help: Option 1 / Option 2

Grant Fritchey SSC Guru Points: 398909 More actions · Answer 3

With the DISTINCT aggregation operation in addition to the GROUP BY operation but no filtering at all, you're going to be somewhat dependent on hardware (mostly memory & disk) to make this run fast. Any chance of some type of WHERE clause? Do you really have to have ALL the data?

"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

ss-457805 SSCertifiable Points: 5872 More actions · Answer 4

Thanks Luis and Chris for your suggestions. Both your query returns in same time. It take about 1 min 15 sec. Big improvement from the orginial query.
Grant - we cannot use the where clause it needs all the data. No wonder it's doing an index scan on on the whole table. Please find attached the execution plan

blog: http://sarveshsingh.com Twitter: @sarveshsing

ChrisM@Work SSC Guru Points: 186127 More actions · Answer 5

ss-457805 - Thursday, July 6, 2017 2:16 PM
Thanks Luis and Chris for your suggestions. Both your query returns in same time. It take about 1 min 15 sec. Big improvement from the orginial query.
Grant - we cannot use the where clause it needs all the data. No wonder it's doing an index scan on on the whole table. Please find attached the execution plan

Thanks for the Execution Plan.
I know this looks a little complicated, but it's extremely fast because it replaces streaming aggregates with seeks, for each supplierid in each of the two big tables.
Useful links here.
Give it a try:
-- Using Paul White's superfast "DISTINCT"

-- 1. Collect all SupplierID from dbo.CPT_SUBDC_Analysis
-- into #temp table #CPT_SUBDC_Analysis_1
WITH rCTE AS (
SELECT SupplierID = MIN(T.SupplierID)
FROM dbo.CPT_SUBDC_Analysis T
UNION ALL
SELECT R.SupplierID
FROM (
  SELECT
   T.SupplierID,
   rn = ROW_NUMBER() OVER (ORDER BY T.SupplierID)
  FROM dbo.CPT_SUBDC_Analysis T
  JOIN rCTE R
   ON R.SupplierID < T.SupplierID
) R
WHERE R.rn = 1
)
SELECT SupplierID
INTO #CPT_SUBDC_Analysis_1
FROM rCTE
OPTION (MAXRECURSION 0);

-- 2. Collect all SupplierID from CPT.dbo.CPT_SUBDC_Analysis
-- into #temp table #CPT_SUBDC_Analysis_2
WITH rCTE AS (
SELECT SupplierID = MIN(T.SupplierID)
FROM CPT.dbo.CPT_SUBDC_Analysis T
UNION ALL
SELECT R.SupplierID
FROM (
  SELECT
   T.SupplierID,
   rn = ROW_NUMBER() OVER (ORDER BY T.SupplierID)
  FROM CPT.dbo.CPT_SUBDC_Analysis T
  JOIN rCTE R
   ON R.SupplierID < T.SupplierID
) R
WHERE R.rn = 1
)
SELECT SupplierID
INTO #CPT_SUBDC_Analysis_2
FROM rCTE
OPTION (MAXRECURSION 0);

-- 3. Use these two temp tables for the query:
SELECT
s.SupplierID,
c.Carrier as Supplier
FROM (
SELECT *
FROM (
  SELECT SupplierID
  FROM #CPT_SUBDC_Analysis_1
  UNION ALL
  SELECT SupplierID
  FROM #CPT_SUBDC_Analysis_2
) d GROUP BY SupplierID
) s
LEFT OUTER JOIN Carrier c
ON c.CarrierID = s.SupplierID
ORDER BY c.Carrier

^{“Write the query the simplest way. If through testing it becomes clear that the performance is inadequate, consider alternative query forms.” - Gail Shaw}

For fast, accurate and documented assistance in answering your questions, please read this article.
Understanding and using APPLY, (I) and (II) Paul White
Hidden RBAR: Triangular Joins / The "Numbers" or "Tally" Table: What it is and how it replaces a loop Jeff Moden

ss-457805 SSCertifiable Points: 5872 More actions · Answer 6

Thanks Chris. This is awesome. It's pretty instant.

Thanks for the link as well. Added to my favourites. 🙂

blog: http://sarveshsingh.com Twitter: @sarveshsing