From: | Tyler Hildebrandt <tyler(at)campbell-lange(dot)net> |
---|---|
To: | pgsql-performance(at)postgresql(dot)org |
Cc: | Sebastian Ritter <sebastian(at)campbell-lange(dot)net>, rory <rory(at)campbell-lange(dot)net> |
Subject: | Query timing increased from 3s to 55s when used as function instead of select |
Date: | 2010-05-21 13:54:15 |
Message-ID: | 20100521135415.GA3445@campbell-lange.net |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-performance |
We're using a function that when run as a select statement outside of the
function takes roughly 1.5s to complete whereas running an identical
query within a function is taking around 55s to complete.
We are lost as to why placing this query within a function as opposed to
substituting the variables in a select statement is so drastically different.
The timings posted here are from a 512MB memory virtual machine and are not of
major concern on their own but we are finding the same issue in our production
environment with far superior hardware.
The function can be found here:
http://campbell-lange.net/media/files/fn_medirota_get_staff_leave_summary.sql
---
Timings for the individual components on their own is as follows:
select * from fn_medirota_validate_rota_master(6);
Time: 0.670 ms
select to_date(EXTRACT (YEAR FROM current_date)::text, 'YYYY');
Time: 0.749 ms
select * from fn_medirota_people_template_generator(2, 6, date'2009-01-01',
date'2009-12-31', TRUE) AS templates;
Time: 68.004 ms
select * from fn_medirota_people_template_generator(2, 6, date'2010-01-01',
date'2010-12-31', TRUE) AS templates;
Time: 1797.323
Copying the exact same for loop select statement from the query above into
the psql query buffer and running them with variable substitution yields the
following:
Running FOR loop SElECT with variable substitution:
Time: 3150.585 ms
Whereas invoking the function yields:
select * from fn_medirota_get_staff_leave_summary(6);
Time: 57375.477 ms
We have tried using explain analyse to update the query optimiser, dropped and
recreated the function and have restarted both the machine and the postgres
server multiple times.
Any help or advice would be greatly appreciated.
Kindest regards,
Tyler Hildebrandt
---
EXPLAIN ANALYSE VERBOSE SELECT * FROM fn_medirota_get_staff_leave_summary(6);
QUERY PLAN
-------------------------------------------------------------------------------------------------------------------------------------------------
{FUNCTIONSCAN
:startup_cost 0.00
:total_cost 260.00
:plan_rows 1000
:plan_width 85
:targetlist (
{TARGETENTRY
:expr
{VAR
:varno 1
:varattno 1
:vartype 23
:vartypmod -1
:varlevelsup 0
:varnoold 1
:varoattno 1
}
:resno 1
:resname id
:ressortgroupref 0
:resorigtbl 0
:resorigcol 0
:resjunk false
}
{TARGETENTRY
:expr
{VAR
:varno 1
:varattno 2
:vartype 1043
:vartypmod -1
:varlevelsup 0
:varnoold 1
:varoattno 2
}
:resno 2
:resname t_full_name
:ressortgroupref 0
:resorigtbl 0
:resorigcol 0
:resjunk false
}
{TARGETENTRY
:expr
{VAR
:varno 1
:varattno 3
:vartype 16
:vartypmod -1
:varlevelsup 0
:varnoold 1
:varoattno 3
}
:resno 3
:resname b_enabled
:ressortgroupref 0
:resorigtbl 0
:resorigcol 0
:resjunk false
}
{TARGETENTRY
:expr
{VAR
:varno 1
:varattno 4
:vartype 1043
:vartypmod -1
:varlevelsup 0
:varnoold 1
:varoattno 4
}
:resno 4
:resname t_anniversary
:ressortgroupref 0
:resorigtbl 0
:resorigcol 0
:resjunk false
}
{TARGETENTRY
:expr
{VAR
:varno 1
:varattno 5
:vartype 23
:vartypmod -1
:varlevelsup 0
:varnoold 1
:varoattno 5
}
:resno 5
:resname n_last_year_annual
:ressortgroupref 0
:resorigtbl 0
:resorigcol 0
:resjunk false
}
{TARGETENTRY
:expr
{VAR
:varno 1
:varattno 6
:vartype 23
:vartypmod -1
:varlevelsup 0
:varnoold 1
:varoattno 6
}
:resno 6
:resname n_last_year_other
:ressortgroupref 0
:resorigtbl 0
:resorigcol 0
:resjunk false
}
{TARGETENTRY
:expr
{VAR
:varno 1
:varattno 7
:vartype 23
:vartypmod -1
:varlevelsup 0
:varnoold 1
:varoattno 7
}
:resno 7
:resname n_this_year_annual
:ressortgroupref 0
:resorigtbl 0
:resorigcol 0
:resjunk false
}
{TARGETENTRY
:expr
{VAR
:varno 1
:varattno 8
:vartype 23
:vartypmod -1
:varlevelsup 0
:varnoold 1
:varoattno 8
}
:resno 8
:resname n_this_year_other
:ressortgroupref 0
:resorigtbl 0
:resorigcol 0
:resjunk false
}
)
:qual <>
:lefttree <>
:righttree <>
:initPlan <>
:extParam (b)
:allParam (b)
:scanrelid 1
:funcexpr
{FUNCEXPR
:funcid 150447
:funcresulttype 149366
:funcretset true
:funcformat 0
:args (
{CONST
:consttype 23
:consttypmod -1
:constlen 4
:constbyval true
:constisnull false
:constvalue 4 [ 6 0 0 0 0 0 0 0 ]
}
)
}
:funccolnames ("id" "t_full_name" "b_enabled" "t_anniversary" "n_last_year_
annual" "n_last_year_other" "n_this_year_annual" "n_this_year_other")
:funccoltypes <>
:funccoltypmods <>
}
Function Scan on fn_medirota_get_staff_leave_summary (cost=0.00..260.00
rows=1000 width=85) (actual time=51877.812..51877.893 rows=94 loops=1)
Total runtime: 51878.008 ms
(183 rows)
--
Tyler Hildebrandt
Software Developer
tyler(at)campbell-lange(dot)net
Campbell-Lange Workshop
www.campbell-lange.net
020 7631 1555
3 Tottenham Street London W1T 2AF
Registered in England No. 04551928
From | Date | Subject | |
---|---|---|---|
Next Message | Matthew Wakeling | 2010-05-21 15:08:38 | Re: Optimize date query for large child tables: GiST or GIN? |
Previous Message | Piotr Legiecki | 2010-05-21 12:15:21 | Re: old server, new server, same performance |