How to speed up this query?

Question

Post reply

How to speed up this query?

water490

SSC Eights!

Points: 918
More actions
July 3, 2023 at 1:04 am
Go to Answer
#4225376
Hi everyone
My query is taking over 5 minutes to run and still no results so I think something is wrong. I am not sure how to fix it. Is someone able to help me?
Query:
```
DROP TABLE IF EXISTS #TEMP1;

SELECT DISTINCT T1.QUOTE_DATE
INTO #TEMP1
FROM DBO.OptionsEOD T1

CREATE CLUSTERED INDEX #CI_TEMP1 ON #TEMP1
(QUOTE_DATE
);

SELECTT1.UNDERLYING_SYMBOL,
T1.QUOTE_DATE,
T1.ROOT_SYMBOL,
T1.EXPIRATION,
T1.STRIKE,
T1.OPTION_TYPE,
T1.CLOSE_PRICE,
LAG(T1.CLOSE_PRICE) OVER (PARTITION BY T1.UNDERLYING_SYMBOL, T1.ROOT_SYMBOL, T1.EXPIRATION, T1.STRIKE, T1.OPTION_TYPE ORDER BY T1.QUOTE_DATE) AS PREV_CLOSE_PRICE
FROMDBO.OptionsEOD AS T1 INNER JOIN
(
SELECTT2.DATE_FROM,
T2.DATE_TO
FROM(
SELECTLAG(T1.QUOTE_DATE,1) OVER (ORDER BY T1.QUOTE_DATE) AS DATE_FROM,
LAG(T1.QUOTE_DATE,0) OVER (ORDER BY T1.QUOTE_DATE) AS DATE_TO,
ROW_NUMBER() OVER (ORDER BY T1.QUOTE_DATE DESC) AS ROW_NUM
FROM#TEMP1 AS T1
) AS T2
WHERET2.ROW_NUM = 1
) AS T3 ON T1.QUOTE_DATE BETWEEN T3.DATE_FROM AND T3.DATE_TO
```
Explanation:
I want to find the previous close price for a security. The composite key are the following fields:
UNDERLYING_SYMBOL
QUOTE_DATE
ROOT_SYMBOL
EXPIRATION
STRIKE
OPTION_TYPE
The LAG function finds the previous value. If I run the query as is without the sub-query then it produces way too many useless results. So the solution is to produce the results for today only. This is financial stock data so there is only data for business days so today (Sunday) means Friday (or Thursday if Friday is a holiday). The sub-query produces the latest day (DATE_TO) and the previous day (DATE_FROM).
The problem is I am not sure how to feed the results of the sub-query to the main query. I used BETWEEN but I am not really sure. There must be a better way.
The main query needs to restrict results so only DATE_FROM and DATE_TO are used. Here is a pseudocode:
Select *
from table
where quote_date <=DATE_TO and quote_date >= DATE_FROM
I don't know how to do this.
How can I fix my query?
Thank you

frederico_fonseca

SSCoach

Points: 16047

July 3, 2023 at 4:23 pm

Answer

#4225981

try the following code.

and then give us the actual explain plan for both this code and your code.

drop table if exists #TEMP1;
drop table if exists #TEMP2;

select distinct T1.QUOTE_DATE
into #TEMP1
from dbo.OptionsEOD T1

CREATE CLUSTERED INDEX #CI_TEMP1 ON #TEMP1
(QUOTE_DATE
);

select T2.DATE_FROM
     , T2.DATE_TO
into #temp2
from (
    select lag(T1.QUOTE_DATE, 1) over (order by T1.QUOTE_DATE) as DATE_FROM
         , lag(T1.QUOTE_DATE, 0) over (order by T1.QUOTE_DATE) as DATE_TO
         , row_number() over (order by T1.QUOTE_DATE desc) as ROW_NUM
    from #TEMP1 as T1) as T2
where T2.ROW_NUM = 1



select T1.UNDERLYING_SYMBOL
     , T1.QUOTE_DATE
     , T1.ROOT_SYMBOL
     , T1.EXPIRATION
     , T1.STRIKE
     , T1.OPTION_TYPE
     , T1.CLOSE_PRICE
     , lag(T1.CLOSE_PRICE) over (partition by T1.UNDERLYING_SYMBOL, T1.ROOT_SYMBOL, T1.EXPIRATION, T1.STRIKE, T1.OPTION_TYPE order by T1.QUOTE_DATE) as PREV_CLOSE_PRICE
from dbo.OptionsEOD as T1
    inner join #temp2 as T3
        on T1.QUOTE_DATE between T3.DATE_FROM and T3.DATE_TO

Viewing 14 posts - 1 through 13 (of 13 total)

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

Neil Burton SSC-Insane Points: 22927 More actions · Answer 1

Have you looked at the query execution plan? There might be some clues as to what is causing the problem in there. What kind of data volumes are you dealing with? A query that takes 5+ minutes over thousands of rows might be a problem but it might be reasonable if you're querying millions of rows. What is the structure of the dbo.OptionsEOD table?

People can probably help you but we're going to need more to go on first.

On two occasions I have been asked, "Pray, Mr. Babbage, if you put into the machine wrong figures, will the right answers come out?" ... I am not able rightly to apprehend the kind of confusion of ideas that could provoke such a question.
—Charles Babbage, Passages from the Life of a Philosopher

How to post a question to get the most help http://www.sqlservercentral.com/articles/Best+Practices/61537

Grant Fritchey SSC Guru Points: 398716 More actions · Answer 2

Nothing jumps out as problematic, so it is going to be down to knowing the structures and execution plan. Without that, we can't even guess.

"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

ScottPletcher SSC Guru Points: 100957 More actions · Answer 3

If you don't already have an index on the dbo.OptionsEOD table to directly support the "LAG(T1.CLOSE_PRICE) OVER PARTITION BY T1.UNDERLYING_SYMBOL, T1.ROOT_SYMBOL, T1.EXPIRATION, T1.STRIKE, T1.OPTION_TYPE ORDER BY T1.QUOTE_DATE)" you should try creating one.

That is, the index keys would be ( UNDERLYING_SYMBOL, ROOT_SYMBOL, EXPIRATION, STRIKE, OPTION_TYPE, QUOTE_DATE ) /* in that order */ and the included column would be CLOSE_PRICE.

SQL DBA,SQL Server MVP(07, 08, 09) A socialist is someone who will give you the shirt off *someone else's* back.

Jonathan AC Roberts SSCoach Points: 18272 More actions · Answer 4

SELECT T1.UNDERLYING_SYMBOL,
       T1.QUOTE_DATE,
       T1.ROOT_SYMBOL,
       T1.EXPIRATION,
       T1.STRIKE,
       T1.OPTION_TYPE,
       T1.CLOSE_PRICE,
       T2.PREV_CLOSE_PRICE
  FROM DBO.OptionsEOD AS T1 
 OUTER APPLY(SELECT TOP(1) 
                    T2.CLOSE_PRICE AS PREV_CLOSE_PRICE
               FROM DBO.OptionsEOD AS T2
              WHERE T2.UNDERLYING_SYMBOL = T1.UNDERLYING_SYMBOL
                AND T2.ROOT_SYMBOL = T1.ROOT_SYMBOL
                AND T2.EXPIRATION  = T1.EXPIRATION
                AND T2.STRIKE = T1.STRIKE
                AND T2.OPTION_TYPE = T1.OPTION_TYPE
                AND T2.QUOTE_DATE < T1.QUOTE_DATE
              ORDER BY T2.QUOTE_DATE  DESC) T2
 WHERE T1.QUOTE_DATE = '2023-07-03'
;

You might need to add an index to make it perform well.

Jonathan AC Roberts SSCoach Points: 18272 More actions · Answer 5

Another option:

WITH CTE AS
(
    SELECT T1.UNDERLYING_SYMBOL,
           T1.QUOTE_DATE,
           T1.ROOT_SYMBOL,
           T1.EXPIRATION,
           T1.STRIKE,
           T1.OPTION_TYPE,
           T1.CLOSE_PRICE,
           LAG(T1.CLOSE_PRICE) OVER (PARTITION BY T1.UNDERLYING_SYMBOL, T1.ROOT_SYMBOL, T1.EXPIRATION, T1.STRIKE, T1.OPTION_TYPE ORDER BY T1.QUOTE_DATE) AS PREV_CLOSE_PRICE
      FROM DBO.OptionsEOD AS T1 
)
SELECT *
  FROM CTE
 WHERE QUOTE_DATE = '2023-07-03'
;

Jonathan AC Roberts SSCoach Points: 18272 More actions · Answer 6

To increase the performance you could limit the rows the query in the CTE produces by selecting on the QUOTE_DATE going back a given amount of days to when you are sure there will be another quote date. This code looks back 10 days.

DECLARE @QUOTE_DATE date = '2023-07-03';

;WITH CTE AS
(
    SELECT T1.UNDERLYING_SYMBOL,
           T1.QUOTE_DATE,
           T1.ROOT_SYMBOL,
           T1.EXPIRATION,
           T1.STRIKE,
           T1.OPTION_TYPE,
           T1.CLOSE_PRICE,
           LAG(T1.CLOSE_PRICE) OVER (PARTITION BY T1.UNDERLYING_SYMBOL, T1.ROOT_SYMBOL, T1.EXPIRATION, T1.STRIKE, T1.OPTION_TYPE ORDER BY T1.QUOTE_DATE) AS PREV_CLOSE_PRICE
      FROM DBO.OptionsEOD AS T1 
     WHERE T1.QUOTE_DATE BETWEEN DATEADD(dd, -10, @QUOTE_DATE) AND @QUOTE_DATE
)
SELECT *
  FROM CTE
 WHERE QUOTE_DATE = @QUOTE_DATE
;

Jonathan AC Roberts SSCoach Points: 18272 More actions · Answer 7

I think the following index should help performance:

 CREATE INDEX IX_OptionsEOD_1 
     ON dbo.OptionsEOD (UNDERLYING_SYMBOL, ROOT_SYMBOL, EXPIRATION, STRIKE, OPTION_TYPE, QUOTE_DATE)
INCLUDE (CLOSE_PRICE)
;

water490 SSC Eights! Points: 918 More actions · Answer 8

frederico_fonseca wrote:

try the following code.

and then give us the actual explain plan for both this code and your code.

drop table if exists #TEMP1;
drop table if exists #TEMP2;

select distinct T1.QUOTE_DATE
into #TEMP1
from dbo.OptionsEOD T1

CREATE CLUSTERED INDEX #CI_TEMP1 ON #TEMP1
(QUOTE_DATE
);

select T2.DATE_FROM
     , T2.DATE_TO
into #temp2
from (
    select lag(T1.QUOTE_DATE, 1) over (order by T1.QUOTE_DATE) as DATE_FROM
         , lag(T1.QUOTE_DATE, 0) over (order by T1.QUOTE_DATE) as DATE_TO
         , row_number() over (order by T1.QUOTE_DATE desc) as ROW_NUM
    from #TEMP1 as T1) as T2
where T2.ROW_NUM = 1



select T1.UNDERLYING_SYMBOL
     , T1.QUOTE_DATE
     , T1.ROOT_SYMBOL
     , T1.EXPIRATION
     , T1.STRIKE
     , T1.OPTION_TYPE
     , T1.CLOSE_PRICE
     , lag(T1.CLOSE_PRICE) over (partition by T1.UNDERLYING_SYMBOL, T1.ROOT_SYMBOL, T1.EXPIRATION, T1.STRIKE, T1.OPTION_TYPE order by T1.QUOTE_DATE) as PREV_CLOSE_PRICE
from dbo.OptionsEOD as T1
    inner join #temp2 as T3
        on T1.QUOTE_DATE between T3.DATE_FROM and T3.DATE_TO

thanks for the reply.

your code:

run time is about 8 seconds

https://www.brentozar.com/pastetheplan/?id=BJ3LCKgY3

my code:

run time is about 7 m 6 s

https://www.brentozar.com/pastetheplan/?id=HyTD15eY3

there is big improvement! is it possible to save even more time?

frederico_fonseca SSCoach Points: 16047 More actions · Answer 9

maybe this but not guaranteed.

drop table if exists #TEMP1;
drop table if exists #TEMP2;
drop table if exists #TEMP3;

select distinct T1.QUOTE_DATE
into #TEMP1
from dbo.OptionsEOD T1

CREATE CLUSTERED INDEX #CI_TEMP1 ON #TEMP1
(QUOTE_DATE
);

select T2.DATE_FROM
     , T2.DATE_TO
into #temp2
from (
    select lag(T1.QUOTE_DATE, 1) over (order by T1.QUOTE_DATE) as DATE_FROM
         , lag(T1.QUOTE_DATE, 0) over (order by T1.QUOTE_DATE) as DATE_TO
         , row_number() over (order by T1.QUOTE_DATE desc) as ROW_NUM
    from #TEMP1 as T1) as T2
where T2.ROW_NUM = 1

select T1.UNDERLYING_SYMBOL
     , T1.QUOTE_DATE
     , T1.ROOT_SYMBOL
     , T1.EXPIRATION
     , T1.STRIKE
     , T1.OPTION_TYPE
     , T1.CLOSE_PRICE
into #temp3
from dbo.OptionsEOD as T1
where 0 = 1;

create clustered index #ci_temp3 on #temp3
(UNDERLYING_SYMBOL
, ROOT_SYMBOL
, EXPIRATION
, STRIKE
, OPTION_TYPE
, QUOTE_DATE
)

insert into #temp3 with (tablock)
select T1.UNDERLYING_SYMBOL
     , T1.QUOTE_DATE
     , T1.ROOT_SYMBOL
     , T1.EXPIRATION
     , T1.STRIKE
     , T1.OPTION_TYPE
     , T1.CLOSE_PRICE
from dbo.OptionsEOD as T1
    inner join #temp2 as T3
        on T1.QUOTE_DATE between T3.DATE_FROM and T3.DATE_TO

select T1.UNDERLYING_SYMBOL
     , T1.QUOTE_DATE
     , T1.ROOT_SYMBOL
     , T1.EXPIRATION
     , T1.STRIKE
     , T1.OPTION_TYPE
     , T1.CLOSE_PRICE
     , lag(T1.CLOSE_PRICE) over (partition by T1.UNDERLYING_SYMBOL, T1.ROOT_SYMBOL, T1.EXPIRATION, T1.STRIKE, T1.OPTION_TYPE order by T1.QUOTE_DATE) as PREV_CLOSE_PRICE
from #temp3 as T1

water490 SSC Eights! Points: 918 More actions · Answer 10

frederico_fonseca wrote:

maybe this but not guaranteed.

drop table if exists #TEMP1;
drop table if exists #TEMP2;
drop table if exists #TEMP3;

select distinct T1.QUOTE_DATE
into #TEMP1
from dbo.OptionsEOD T1

CREATE CLUSTERED INDEX #CI_TEMP1 ON #TEMP1
(QUOTE_DATE
);

select T2.DATE_FROM
     , T2.DATE_TO
into #temp2
from (
    select lag(T1.QUOTE_DATE, 1) over (order by T1.QUOTE_DATE) as DATE_FROM
         , lag(T1.QUOTE_DATE, 0) over (order by T1.QUOTE_DATE) as DATE_TO
         , row_number() over (order by T1.QUOTE_DATE desc) as ROW_NUM
    from #TEMP1 as T1) as T2
where T2.ROW_NUM = 1

select T1.UNDERLYING_SYMBOL
     , T1.QUOTE_DATE
     , T1.ROOT_SYMBOL
     , T1.EXPIRATION
     , T1.STRIKE
     , T1.OPTION_TYPE
     , T1.CLOSE_PRICE
into #temp3
from dbo.OptionsEOD as T1
where 0 = 1;

create clustered index #ci_temp3 on #temp3
(UNDERLYING_SYMBOL
, ROOT_SYMBOL
, EXPIRATION
, STRIKE
, OPTION_TYPE
, QUOTE_DATE
)

insert into #temp3 with (tablock)
select T1.UNDERLYING_SYMBOL
     , T1.QUOTE_DATE
     , T1.ROOT_SYMBOL
     , T1.EXPIRATION
     , T1.STRIKE
     , T1.OPTION_TYPE
     , T1.CLOSE_PRICE
from dbo.OptionsEOD as T1
    inner join #temp2 as T3
        on T1.QUOTE_DATE between T3.DATE_FROM and T3.DATE_TO

select T1.UNDERLYING_SYMBOL
     , T1.QUOTE_DATE
     , T1.ROOT_SYMBOL
     , T1.EXPIRATION
     , T1.STRIKE
     , T1.OPTION_TYPE
     , T1.CLOSE_PRICE
     , lag(T1.CLOSE_PRICE) over (partition by T1.UNDERLYING_SYMBOL, T1.ROOT_SYMBOL, T1.EXPIRATION, T1.STRIKE, T1.OPTION_TYPE order by T1.QUOTE_DATE) as PREV_CLOSE_PRICE
from #temp3 as T1

thank you.

it takes the same time as before...8s

https://www.brentozar.com/pastetheplan/?id=SkCHR0lK2

DesNorton SSC-Insane Points: 24288 More actions · Answer 11

From the plan, it appears that you are returning 5.3mil rows.

Do you REALLY need that much data in the results?

How to post data/code on a forum to get the best help.

frederico_fonseca SSCoach Points: 16047 More actions · Answer 12

5.3 million rows is the estimated - the actual is 37k rows

Statistics on columnstore can be a bit tricky as they don't quite work as other tables - but I would not be surprised that this table is subject to update/deletes (very bad on columnstore) and/or stats not updated frequently.