DOAG: Visual SQL Tuning

Visual SQL Tuning (VST)

Kyle Hailey
kylehailey.com
kyle@delphix.com

Who is Kyle Hailey
 1990 Oracle
–
–
–
–
–

90 support
92 Ported v6
93 France
95 Benchmarking
98 ST Real World Performance

 2000 Dot.Com
 2001 Quest
 2002 Oracle OEM 10g

Success!
First successful OEM design

Who is Kyle Hailey
 1990 Oracle
–
–
–
–
–






90 support
92 Ported v6
93 France
95 Benchmarking

2000 Dot.Com
2001 Quest
2002 Oracle OEM 10g
2005 Embarcadero
– DB Optimizer

Who is Kyle Hailey
• 1990 Oracle






90 support
92 Ported v6
93 France
95 Benchmarking

•
•
•
•

2000 Dot.Com
2001 Quest
2002 Oracle OEM 10g
2005 Embarcadero
 DB Optimizer
• Delphix
When not being a Geek
- Have a little 4 year old boy who takes up all my time

Production
Instance

Database

File system
File system

QA

UAT

Instance

Instance

Instance

Database

Database

Database

File system
File system

File system
File system
File system

File system
File system

Development

Production
Instance

Database

File system

Development

QA

UAT

Instance

Instance

Instance

Database

Database

Database

SQL Tuning

Methodology
1. Find: Problem SQL
2. Study: SQL Execution Plan
3. Fix: How ?

Step 1 : Find “bad” SQL
1. Users complain
2. High Resources
3. Monitoring

Step 2: Get Explain Plan

Trace fie

PARSING IN CURSOR #2 len=53 dep=0 uid=61 oct=3
-------------------------------------------------------------------------------------------lid=61 tim=1151519905950403 hv=2296704914
| Id | Operation
| Name
| Starts | E-Rows | A-Rows |
--------------------------------------------------------------------------------------------ad='4e50010c'
|
1 || HASH GROUP BY
|
|
1 |
1 |
1 |
SELECT 'Hello, world; today is ' | SYSDATE FROM dual
|* 2 |
FILTER
|
|
1 |
|
1909 |
END OF STMT
|* 3 |
TABLE ACCESS BY INDEX ROWID
| PS_RETROPAYPGM_TBL |
1 |
1 |
3413 |
PARSE
|
4 |
NESTED LOOPS
|
|
1 |
165 |
6827 |
|* 5 |
HASH JOIN
|
|
1 |
165 |
3413 |
#2:c=4000,e=1540,p=0,cr=0,cu=0,mis=1,r=0,dep=0,og=1,
HASH JOIN
|
|
1 |
165 |
3624 |
tim=1151519905950397 |* 6 |
|
7 |
TABLE ACCESS BY INDEX ROWID | WB_JOB
|
1 |
242 |
2895 |
BINDS #2:
|
8 |
NESTED LOOPS
|
|
1 |
233 |
2897 |
EXEC
|
9 |
TABLE ACCESS BY INDEX ROWID| PS_PAY_CALENDAR
|
1 |
1 |
1 |
|* 10 |
INDEX RANGE
| PS0PAY_CALENDAR
|
1 |
1 |
1 |
#2:c=0,e=58,p=0,cr=0,cu=0,mis=0,r=0,dep=0,og=1,tim=1 SCAN
|* 11 |
INDEX RANGE SCAN
| WBBJOB_B
|
1 |
286 |
2895 |
151519906034782
|
TABLE
FULL
| WB_RETROPAY_EARNS |
1 | 27456 |
122K|
WAIT #2: nam='SQL*Net |* 12 | to client' ela=ACCESSFULL
message
2 driver
| 13
TABLE ACCESS
| PS_RETROPAY_RQST
|
1 | 13679 | 13679 |
id=1650815232 #bytes=1 |* 14 obj#=-1 INDEX RANGE SCAN
p3=0 |
| PS#RETROPAYPGM_TBL |
3413 |
1 |
3413 |
SORT AGGREGATE
|
|
1791 |
1 |
1791 |
tim=1151519906034809 | 15 |
| 16 |
FIRST ROW
|
|
1791 |
1 |
1579 |
FETCH
|* 17 |
INDEX RANGE SCAN (MIN/MAX)
| WB_JOB_F
|
1791 |
1 |
1579 |
#2:c=0,e=29,p=0,cr=0,cu=0,mis=0,r=1,dep=0,og=1,tim=1
| 18 |
SORT AGGREGATE
|
|
1539 |
1 |
1539 |
151519906034864
| 19 |
FIRST ROW
|
|
1539 |
1 |
1539 |
INDEX RANGE
|
1539 |
1 |
1539 |
WAIT #2: nam='SQL*Net |* 20 | from client' ela= 215 SCAN (MIN/MAX) | WB_JOB_G
message
--------------------------------------------------------------------------------------------driver id=1650815232 #bytes=1 p3=0 obj#=-1
tim=1151519906035133
FETCH
Predicate Information (identified by operation id):
--------------------------------------------------#2:c=0,e=1,p=0,cr=0,cu=0,mis=0,r=0,dep=0,og=0,tim=11
51519906035165
2 - filter(("B"."EFFDT"= AND
WAIT #2: nam='SQL*Net message to client' ela= 1 driver "B"."EFFSEQ"=))
3 - filter("E"."OFF_CYCLE"="A"."PAY_OFF_CYCLE_CAL")
id=1650815232 #bytes=1 p3=0-obj#=-1
5
access("D"."RETROPAY_SEQ_NO"="C"."RETROPAY_SEQ_NO")
6 - access("C"."EMPLID"="B"."EMPLID" AND "C"."EMPL_RCD#"="B"."EMPL_RCD#")
tim=1151519906035188
10 - access("A"."RUN_ID"='PD2' AND "A"."PAY_CONFIRM_RUN"='N')
WAIT #2: nam='SQL*Net message from client' ela= 192
11 - access("B"."COMPANY"="A"."COMPANY" AND "B"."PAYGROUP"="A"."PAYGROUP")
driver id=1650815232 #bytes=1 p3=0 obj#=-1
12 - filter(("C"."RETROPAY_PRCS_FLAG"='C' AND "C"."RETROPAY_LOAD_SW"='Y'))
tim=1151519906035400
14 - access("E"."RETROPAY_PGM_ID"="D"."RETROPAY_PGM_ID")
17 - obj=0 op='FAST
STAT #2 id=1 cnt=1 pid=0 pos=1access("F"."EMPLID"=:B1 AND "F"."EMPL_RCD#"=:B2 AND "F"."EFFDT"<=:B3)
20
access("G"."EMPLID"=:B1 AND "G"."EMPL_RCD#"=:B2 AND "G"."EFFDT"=:B3)
DUAL (cr=0 pr=0 pw=0 time=3-us)

Step 3: ???
Methodology

• Identify Slow Queries
• Look at Execution Plan
• Fix
WTF??

Fix
1. Analyze stats
2. Go to step 1

Explain Plan: Like Directions

Where is the Map?

SQL Tuning

1. Two Table Join
2. Multi-Table Join
3. Create Map (VST diagram)
4. Apply Methodology (how tune)
5. Examples

How to Join two tables?

1. Which table to start with (main step)
2. What Indexes to use
3. What type of join to use HJ, NL
Design options:

•
•
•
•
•
•

Partitions
IOT
Bitmap indexes
Hash cluster
Materialized views
etc

Two Table Join

1. Do Indexes exist
2. What Type of Relation between tables
–
–
–

One to One
One to Many
Many to Many

3. Special Cases
–
–

Outer Joins
Not Exists

Table Join Order
select
from
where

*
a, b
a.id = b.id

If No Index then (NL) order doesn‟t matter
A
id

B
data

id

data

Every row visits Every row
Work= A-rows x B-rows (8 x 4 )
HJ could optimizes => simulate index lookup

2 Table join, with indexes
select
from
where

Indexes on

*
a, b

• a.id
• b.id

a.id = b.id

A

B

B

A

select
from
where

Indexes on

*
a, b

• a.id
• b.id

a.id = b.id
B

A
data

B

id

8

id

data

A

select
from
where

Indexes on

*
a, b

• a.id
• b.id

a.id = b.id
B

A
data

A

B

id

8

id

data

data

id

With Indexes
Start with B

Start with the least rows

4

id

data

2 table join - filters
select
from
where
and
and

A

data

2

*
a, b
a.id = b.id
a.field = „val a‟
b.field = „val b‟

-- join
-- filter a
-- filter b

id
id

data

3 B
Start on table with least rows after filter

Two Table Summary
1. If no indexes & no filters
then order doesn’t matter with NL
• Hash join will simulate index, start by hashing smaller table

2. If indexes
then start with smaller table
3. If filters and indexes
then start with most filtered table

Creating Two table Join Map
• Set diagrams
• Short comings of set diagrams
• New map diagrams

Set Diagrams: Two Table Joins – looks
simple

A

B

Inner Join

(blue is data returned)

Two Table Joins
A

B

Left join B B(+)

A

B

Inner Join

A

B

Not exists B

Two Table Joins
A

B

Left join B B(+)

A

A

B

Right join B A(+)

B

Inner Join

A

B

Not exists B

A

B

Not Exists A

Two Table Joins (blue is data returned)
A

Left join B B(+)

A

B

Inner Join

A

A

B

B

Full outer (union)

B

Right join B A(+)

Doesn’t tell the whole
story
Missing « amplifications »

B
A

Not exists B

A

B

Union of Not Exists

A

B

Not Exists A

One to One - intersection
A

B
1

1

2

2

A

1

1

2

2

A
MAP

B

B

One to Many - projection
1

1

1

A

2

2
2

B

A

1

1

1

1

2

2

2

2

A

MAP

B

B

Many to Many - amplification
A
1

1

1

B

1

1
1

A
rA*rB
------------------------min(ndv(A),ndv(B))

B
1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

1

B

A
MAP

Two Table Joins
A

B

Left join B B(+)

A
MAP

o

B

A

B

Right join B A(+)

A

o

B
MAP

Outer Join – Attention when no columns
selectedthat return no data
Color nodes
o
A

B
no affect

A

V

Left join B B(+)

o

B

Select a.* from a, b where b.field(+) = a.field;

Outer Join – Attention when no columns
selectedthat return no data
Color nodes
o
A

B
no affect

Left join B B(+)

o

V

A

B

Select a.* from a, b where b.field(+) = a.field;

V

o

?

V

A

B

Multiplier

no affect
or
multiplies rows

Exists ( in)
SELECT d.*
FROM dept d WHERE exists
( SELECT null
FROM emp e
WHERE e.deptno=d.deptno);
SELECT d.*
FROM dept d WHERE d.deptno in
( SELECT deptno
FROM emp e );

dept

MAP

E

emp

Not Exists ( not in)
SELECT d.*
FROM dept d WHERE not exists
( SELECT null
FROM emp e
WHERE e.deptno=d.deptno);
SELECT d.*
FROM dept d WHERE d.deptno not in
( SELECT deptno
FROM emp e
where deptno is not null );

dept

MAP

N

emp

Two Table Joins : Maps
A

B

A

o

B

Left join B B(+)

A
A

B

Inner Join

A

B

Not exists B

B

A

B

A

B

A

N

B

Two Table Joins : maps
A

B

A

o

B

A

o

B

A
B

Inner Join

A

B

Not exists B

B

Right join B A(+)

Left join B B(+)

A

A

B

A

B

A

B
N

A

N

B

A

N

B

A

B

Not Exists A

Two Table Join

1. Where to start
1. Start at table with the least rows after predicate filtering

2. Drawing Map:
– One to One / One to Many / Many to Many
B

A

–

B

A

A

Outer Joins / Not Exists
A

o

B

A

Now , three table joins:

N

B

B

Three table join
order_lines

orders

customer

Join direction

8 rows

4 rows

1 row
Ex) 1 customer 4 orders, each order has 2 order_lines each

Join direction
1 row

1 row

1 row

Three table join
order_lines

orders

F
Where
order_lines.field = value

A Filter on order_lines is going to eliminate work
on orders and customers, if we start at order_lines

customer

Three table join
order_lines

orders

customer

F
Where
order_lines.field = value

Starting with a filter on order_lines is going to eliminate work

Three table join
order_lines

orders

customer

2
F

1

Where
orders.field = value
Starting with a Filter on orders is going to eliminate work
Then Join to customers => keeps the number of rows the same

Three table join
order_lines

orders

customer

F

Where
customer.field = value
Starting on a filter on customer is going to eliminate work

Three table join
order_lines

orders

F

F

customer

F

What if a filter on all three?
Choose the one that filter‟s the highest
percent of the table.
100% * (select count(*) from TAB where condition)
--------------------------(select count(*) from Tab)

4 or more table join

Yes

No

Oracle can only join one table in at a time.
Oracle can’t (normally) join two different sub-branches

Put it all together : VST

1. Tables
– drawn as nodes

2. Joins :
– drawn map connector lines
• One-to-one, one-to-many, many-to-many
• Exists, not exists, outer joins

3. Filters
– mark on each table with filter in where clause
51

11/25/2013

How to VST: Tables and Joins
SELECT C.Phone_Number, C.Honorific, C.First_Name, C.Last_Name,
Tables
C.Suffix, C.Address_ID, A.Address_ID, A.Street_Address_Line1, Orders O,
A.Street_Address_Line2, A.City_Name, A.State_Abbreviation,
Order_Details OD,
A.ZIP_Code, OD.Deferred_Shipment_Date, OD.Item_Count,
Products P,
ODT.Text, OT.Text, P.Product_Description, S.Shipment_Date
Customers C,
FROM Orders O, Order_Details OD, Products P, Customers C, Shipments S,
Addresses A, Code_Translations ODT, Code_Translations OT
Shipments S,
WHERE UPPER(C.Last_Name) LIKE :Last_Name||'%'
Addresses A,
AND UPPER(C.First_Name) LIKE :First_Name||'%'
Code_Translations ODT,
AND OD.Order_ID = O.Order_ID
Code_Translations OT
AND O.Customer_ID = C.Customer_ID
AND OD.Product_ID = P.Product_ID(+)
Joins
AND OD.Shipment_ID = S.Shipment_ID(+)
OD.Order_ID = O.Order_ID
AND S.Address_ID = A.Address_ID(+)
AND O.Status_Code = OT.Code
O.Customer_ID = C.Customer_ID
AND OD.Status_Code = ODT.Code
OD.Product_ID = P.Product_ID(+)
AND O.Order_Date > :Now - 366
OD.Shipment_ID =
ORDER BY C.Customer_ID, O.Order_ID DESC, S.Shipment_ID, OD.Order_Detail_ID;

S.Shipment_ID(+)
S.Address_ID = A.Address_ID(+)
O.Status_Code = OT.Code
OD.Status_Code = ODT.Code

Filters
AND UPPER(C.First_Name) LIKE :First_Name||'%’
AND O.Order_Date > :Now – 366

Dan Tow – SQL TUNING

Layout tables and connections
Tables
Orders O,
Order_Details OD,
Products P,
Customers C,
Shipments S,
Addresses A,
Code_Translations ODT,
Code_Translations OT

O

S

OD
A

P

C
Joins
OD.Shipment_ID = S.Shipment_ID(+)

ODT
OT

F

Dan Tow – SQL TUNING

Unstructured
Joins

OT
A
C

O

S

OD
P

ODT

Neater, but can you do anything with it?
What‟s the optimal execution path?

Copyright 2006

Parents and Children
Structure
the
tree

No index or non-unique

Detail

Primary Key (unique index)

Master

Joins

OD

S

A

P

O

ODT

C

OT

VST – filters and best path
 Filters help determine best path
OD

Filters
AND UPPER(C.First_Name) LIKE :First_Name||'%’
AND O.Order_Date > :Now – 366

S

P

F O

ODT

30%

F C

A

OT

0.02%

100% * (select count(*) from TAB where condition)
--------------------------(select count(*) from Tab)

Child
Concept:
1. Start at most selective filter
2. Join down first, before joining upwards

Child

Parent

Parent

VST – best path

OD

S

P

F O
F

ODT
30%

A

F C
F
0.02%

OT

VST – best path OD

S

P

F O
F

ODT
30%

A

F C
F

OT

0.02%

/*+ Leading (C,O, OT, OD ) */

Dan Tow
SQL TUNING

M

T1
.3

T4

T9
.6

T2

T5

T3

T7
.001

T6

T10
.5

T11

T15
.8

T12
.5
T16

T8
.7
T13
.2

T14
.9

M

T1
.3

T4

T9
.6

T2

T5

T7
.001

T6

T10
.5

T11

T15
.8
T7 -> T13

T3

T12
.5
T16

T8
.7
T13
.2

T14
.9

M

T1
.3

T4

T9
.6

T2

T5

T6

T10
.5

T7 ->T13

T11

T15
.8
T7 -> T13

T3

T12
.5
T16

T8
.7
T14
.9

M

T1
.3

T4

T9
.6

T2

T5

T6

T10
.5

T7 -> T13

T7 ->T13
-> T12

T11

T15
.8

T3

T8
.7
T14
.9

T16

M
T7 ->T13
-> T12 ->
T2

T1
.3

T4

T9
.6

T5

T6

T10
.5

T7 -> T13

T8
.7

T11

T15
.8

T3

T14
.9
T16

M

T1
.3

T3
T7, T13, T12
, T2 , T6

T4

T9
.6

T5

T8
.7

T10
.5

T11

T15
.8
T7 -> T13

T14
.9
T16

M

T1
.3

T3
T7, T13, T12
, T2 , T6, T10

T4

T5

T8
.7

T9
.6

T11

T15
.8
T7 -> T13

T14
.9
T16

M

T1
.3

T4

T3

T7, T13, T12
, T2 , T6,
T10, T11

T5

T9
.6

T14
.9
T15
.8

T7 -> T13

T8
.7

T16

M

T1
.3

T4

T5

T3
T7, T13, T12
, T2 , T6,
T10, T11,
T15

T9
.6

T14
.9
T16

T7 -> T13

T8
.7

M

T1
.3

T4

T9
.6

T7 -> T13

T5

T3
T7, T13, T12
, T2 , T6,
T10, T11,
T15, T16

T8
.7
T14
.9

T1
.3

T4

T9
.6

T7 -> T13

T5

T7, T13, T12
, T2 , T6,
T10, T11,
T15, T16, M

T3

T8
.7
T14
.9

T4

T9
.6

T7 -> T13

T5

T7, T13, T12
, T2 , T6,
T10, T11,
T15, T16, M,
T1

T3

T8
.7
T14
.9

T7, T13, T12
, T2 , T6,
T10, T11,
T15, T16, M,
T1, T4
T5
T9
.6

T7 -> T13

T3

T8
.7
T14
.9

T7, T13, T12
, T2 , T6,
T10, T11,
T15, T16, M,
T1, T4,T9
T5

T3

T8
.7
T14
.9

T7 -> T13

T5

T7 -> T13

T7, T13, T12
, T2 , T6,
T10, T11,
T15, T16, M,
T1,
T4,T9,T3,T8T
14

T7, T13, T12
, T2 , T6,
T10, T11,
T15, T16, M,
T1,
T4,T9,T3,T8T
14, T5

Why not Guess ?
For 17 table join, there are 355 Trillion combinations

Cartesian
SELECT
A.BROKER_ID BROKER_ID,
A.BROKER_LAST_NAME BROKER_LAST_NAME,
A.BROKER_FIRST_NAME BROKER_FIRST_NAME,
A.YEARS_WITH_FIRM YEARS_WITH_FIRM,
C.OFFICE_NAME OFFICE_NAME,
SUM (B.BROKER_COMMISSION)
TOTAL_COMMISSIONS
FROM
BROKER A,
CLIENT_TRANSACTION B,
OFFICE_LOCATION C,
INVESTMENT I
WHERE
A.BROKER_ID = B.BROKER_ID AND
A.OFFICE_LOCATION_ID =
C.OFFICE_LOCATION_ID
GROUP BY
A.BROKER_ID,
A.BROKER_LAST_NAME,
A.BROKER_FIRST_NAME,
A.YEARS_WITH_FIRM,
C.OFFICE_NAME;

Implied Cartesian
select
c.client_first_name, c.client_last_name,
ct.action, ct.price,
b.broker_last_name, b.broker_first_name,
o.office_name
from
client_transaction ct,
client c,
broker b,
office_location o
where
ct.price > 100
and b.broker_id=ct.broker_id
and c.broker_id = b.broker_id
and o.office_location_id = b.office_location_id

Diagram work for Many to One

What about many to many?

Unstructured
OD

Joins

S

P

ODT

C

A

O

OT

OT
A
C
O

S

OD
P

ODT

Copyright 2006

Many-to-One vs Many-to-Many
3

1
Predicate Filter

Now what?
2

B -> C -> A

go to A or C?

Adding Constraints
SELECT COUNT (*)
FROM
b,
c,
a
WHERE
b.val2 = 100 AND
a.val1 = b.id AND
b.val1 = c.id;

58 logical reads

alter table c add constraint c_pk unique (id);
alter table b add constraint b_pk unique (id);

7 logical reads

Diagraming: what to do with many to many
• Problem : with many-to-many don’t know where do we go

• Solution: Add two table join result set sizes

Look at 3 queries
• Query 1 runs more than 24 hours
• Query 2 outer joins and scalar sub-queries
• Query 3 create path not available to Oracle

Query 1 : Over 24 hours to run
SELECT
A0.zuchinis,
A0.brocoli,
C0.Oranges
FROM
(
SELECT
A1.planted_date,
A1.pears,
A1.zuchinis,
A1.brocoli
FROM
FOO.A A1,
(
SELECT
zuchinis,
brocoli
FROM FOO.A A2
WHERE
pears = 'M' AND
planted_date + 0 >= ADD_MONTHS ((SELECT
MAX (planted_date)
FROM FOO.B B1
WHERE
pears = 'M'
),
- 11)
GROUP BY
zuchinis,
brocoli
HAVING COUNT (*) = 12
)
i2
WHERE
A1.planted_date = (SELECT
MAX (planted_date)
FROM FOO.B B2
WHERE
pears = 'M'
) AND
A1.pears = 'M' AND
A1.zuchinis = i2.zuchinis (+) AND
A1.brocoli = i2.brocoli (+)
UNION
SELECT
A4.planted_date,

A4.pears,A4.zuchinis,A4.brocoli
FROM FOO.A A4
WHERE
A4.planted_date >'01-OCT-08' and A4.planted_date <'03-OCT-08' AND
A4.pears = 'D' AND
A4.green_beans = '1'
AND NOT EXISTS (SELECT
*
FROM FOO.A A5
WHERE
pears = 'M' AND
planted_date = (SELECT
MAX (planted_date)
FROM FOO.B B3
WHERE
pears = 'M'
) AND
A4.zuchinis = A5.zuchinis AND
A4.brocoli = A5.brocoli)
)
b,
FOO.A A0,
FOO.C C0,
FOO.D D0,
FOO.E E0
WHERE
A0.planted_date >'01-OCT-08' and
A0.planted_date <'03-OCT-08' AND
A0.pears = 'D' AND
A0.green_beans = '1' AND
A0.zuchinis = b.zuchinis AND
A0.brocoli = b.brocoli AND
A0.planted_date = C0.planted_date AND
A0.pears = C0.pears AND
A0.zuchinis = C0.zuchinis AND
A0.brocoli = C0.brocoli AND
A0.planted_date = D0.planted_date AND
A0.pears = D0.pears AND
A0.harvest_size = D0.harvest_size AND
C0.Oranges = D0.Oranges AND
C0.apples = D0.apples AND
(D0.lemons = 0 OR
D0.lemons IS NULL) AND
A0.planted_date = E0.planted_date AND
A0.pears = E0.pears AND
A0.harvest_size = E0.harvest_size AND
C0.Oranges = E0.Oranges AND
C0.apples = E0.apples AND
(E0.lemons = 0 OR
E0.lemons IS NULL)
ORDER BY
A0.zuchinis, A0.brocoli;

Default vs Tuned
Default

Tuned

Comparing Plans : 24 hours to 5 mins

NL
NL

NL

NL

HJ
NL
HJ
NL

Q2
SELECT CASE WHEN M.NYC IS NULL THEN (SELECT /*+ qb_name(qb1) */ MAX (Kona)
FROM foo.F
WHERE harvest_date = to_date('08/10/2008','dd/mm/yyyy')
AND Argentina = TRIM ('D') AND Norway = F_OUTER.Norway
ELSE M.NYC END AS NYC,
CASE WHEN F_OUTER.Perth IS NULL THEN NULL
ELSE (SELECT /*+ qb_name(qb2) */ Georgia FROM foo.P
AND Argentina = TRIM ('D') AND Paris = F_OUTER.Perth)
END AS richard,
CASE WHEN F_OUTER.Aruba IS NULL THEN NULL
AND Argentina = TRIM ('D') AND Paris = F_OUTER.Aruba)
END AS Jody,
CASE WHEN F_OUTER.Portland IS NULL THEN NULL
AND Argentina = TRIM ('D') AND Paris = F_OUTER.Portland)
END AS Tom
FROM foo.F F_OUTER, foo.M , foo.J , foo.N ,
(SELECT /*+ qb_name(qb5) */ H.SF, Oregon, H.Haiti, K.Bermuda, L.Denmark
FROM (foo.H LEFT OUTER JOIN foo.K
ON H.harvest_date = K.harvest_date
AND H.Argentina = K.Argentina AND H.SF = K.SF
AND K.Dallas = '001')
LEFT OUTER JOIN FOo.L
ON H.harvest_date = L.harvest_date
AND H.Argentina = L.Argentina AND H.SF = L.SF
WHERE H.harvest_date = to_date('08/10/2008','dd/mm/yyyy')
AND H.Argentina = TRIM ('D')) extra
WHERE F_OUTER.harvest_date = M.harvest_date(+)
AND F_OUTER.Argentina = M.Argentina(+)
AND F_OUTER.Norway = M.Norway(+)
AND M.Norway(+) = M.Texas(+)
AND F_OUTER.harvest_date = to_date('08/10/2008','dd/mm/yyyy')
AND F_OUTER.Argentina = TRIM ('D')
AND M.harvest_date(+) = to_date('08/10/2008','dd/mm/yyyy')
AND M.Argentina(+) = TRIM ('D')
AND F_OUTER.Norway = F_OUTER.Hawaii
AND F_OUTER.harvest_date = J.harvest_date(+)
AND F_OUTER.Argentina = J.Argentina(+)
AND F_OUTER.Norway = J.Texas(+)
AND J.harvest_date(+) = to_date('08/10/2008','dd/mm/yyyy')
AND J.Argentina(+) = TRIM ('D')
AND F_OUTER.Iraq = extra.SF(+)
AND F_OUTER.harvest_date = N.harvest_date(+)
AND F_OUTER.Argentina = N.Argentina(+)
AND F_OUTER.Norway = N.Hawaii(+)
AND N.Jordon(+) = '0'
/

4 Scalar sub-queries
In the select clause

Scalar Sub-queries
682348

845
825

12

Q2
The subqueries in the select clause look like
select CASE WHEN F.f1 IS NULL
THEN NULL
THEN NULL
ELSE (SELECT X.f2
ELSE (SELECT X.f2
FROM X
FROM X
WHERE code_vl = F.f1)
WHERE code_vl = F.f2)
END AS f0 f0
END AS
from F;
from F;

and should be merged into the query like:
THEN NULL
THEN NULL
ELSE ( X.f2)
ELSE ( X.f2)
END f0
END ASAS f0
from F , X
from F , X
where code_vl(+) = decode(f.f1, null, null, F.F2)
where code_vl(+) = F.f1;

Q3
SELECT DISTINCT *
FROM
FOO.a a, FOO.c c, FOO.d d, FOO.g g
WHERE
a.planted_date > '01-OCT-08' AND
a.planted_date < '03-OCT-08' AND
a.pears = 'D' AND a.green_beans = '1' AND
a.planted_date = c.planted_date AND
a.pears = c.pears AND
a.zuchinis = c.zuchinis AND
a.brocoli = c.brocoli AND
a.planted_date = d.planted_date AND
a.pears = d.pears AND
a.harvest_size = d.harvest_size AND
c.oranges = d.oranges AND
c.apples = d.apples AND
(d.lemons = 0 OR d.lemons IS NULL) AND
a.planted_date = g.planted_date AND
a.pears = g.pears AND
a.harvest_size = g.harvest_size AND
c.oranges = g.oranges AND
c.apples = g.apples AND
(g.lemons = 0 OR g.lemons IS NULL) AND
a.zuchinis = '0236' AND
d.apples = g.apples AND
d.oranges = g.oranges
ORDER BY a.zuchinis, a.brocoli;

65438157

Q3: Transitivity

1,126,402
65438157

7,136,362

95

11/25/2013

Q3
SELECT * FROM
(
SELECT /*+ NO_MERGE */ c.apples, c.oranges, a.harvest_size
FROM a, c
WHERE
a.planted_date = TO_DATE ('02/10/2008', 'dd/mm/yyyy') AND
a.pears = 'D' AND
a.green_beans = '1' AND
a.planted_date = c.planted_date AND
a.pears = c.pears AND
a.zuchinis = c.zuchinis AND
a.brocoli = c.brocoli AND
a.zuchinis = '0236'
) X,
(
SELECT /*+ NO_MERGE */ d.apples, d.oranges, d.harvest_size
FROM d, g
WHERE
d.planted_date = TO_DATE ('02/10/2008', 'dd/mm/yyyy') AND
g.planted_date = TO_DATE ('02/10/2008', 'dd/mm/yyyy') AND
g.apples = d.apples AND
d.oranges = g.oranges AND
d.pears = 'D' AND
g.pears = 'D' AND
g.pears = d.pears AND
g.harvest_size = d.harvest_size AND
(d.lemons = 0 OR d.lemons IS NULL) AND
(g.lemons = 0 OR g.lemons IS NULL)
)Y
WHERE
X.oranges = Y.oranges AND
X.apples = Y.apples AND
X.harvest_size = Y.harvest_size;

65438157

This final version runs in
elapsed 0.33 secs and 12K logical reads
down from an original
elapsed 4.5 secs and 1M logical reads

96

11/25/2013

Related info
• Hints
– Leading hint
– HJ & NL

• Tools to use
– graphical execution = visual SQL tuning (VST) diagram

• Example from Jonathan Lewis
• Join Ratios
– Join detail ratio
– Join master ratio

Leading Hint, use table alias (not table)

OD

S

P

/*+ Leading (C,O, OT, OD ) */

F O
F

ODT
30%

A

F C
F
0.02%

OT

HINTS
•
•
•
•
•

Leading (tab_alias , table_alias … ) 10g (9i use ORDERED, not as good )
USE_NL (table_alias) – 2nd in xplan probed into (Inner Table)
USE_HASH (table_alias) – 1st in xplan probed into
INDEX (tab_alias index_name)
NO_MERGE

Oracle first decides join order then join type

(example http://www.adp-gmbh.ch/blog/2008/01/17.php)

99

11/25/2013

select

/*+ LEADING(X,Y) USE_NL(Y) */ *

NL
X
Y

X

select

HJ
X
Y

Y
/*+ LEADING(X,Y) USE_HASH(Y) */ *

Y

Hash X
results

What tools to use to create VST diagrams?
• Paper
– Modifications messy and difficult

• PowerPoint
– Can move things around
– Sticky connectors
– Easy to modify

• DB Optimizer
– Automatic and still modifiable

VST vs Explain Plan
SELECT
O.ORDER_ID,
LINE_ITEM_ID,
PRODUCT_ID,
UNIT_PRICE,
QUANTITY,
ORDER_MODE,
ORDER_STATUS,
ORDER_TOTAL,
SALES_REP_ID,
PROMOTION_ID,
C.CUSTOMER_ID,
CUST_FIRST_NAME,
CUST_LAST_NAME,
CREDIT_LIMIT,
CUST_EMAIL,
ORDER_DATE
FROM
ORDERS O,
ORDER_ITEMS OI,
MJ
CUSTOMERS C
WHERE
O.ORDER_ID = OI.ORDER_ID AND
O.CUSTOMER_ID = C.CUSTOMER_ID AND
O.ORDER_STATUS <= 4

F

F

Table Sizes
Join Sizes
Filter Ratios


Exists

Exists Subquery

Join Filter ratios

Detail

A
Jd “Join detail ratio”
= avg #rows returned by joining one row from master

Master

Jm “Join master ratio”
= avg #rows returned by joining one row from Detail

Join Filter ratios
FK

Detail

A
Jd > 1

Join Detail Jd > 1 normally

If Jd < 1 joining acts like a filter, unusual
If Jd = 0 then good place to start, rare

PK
Master

Join Filter ratios
FK

Detail
A

Join Detail Jd > 1 normally
Jd > 1
If Jd < 1 joining acts like a filter, unusual
If Jd = 0 then good place to start, rare

PK
Master
Detail

Detail
Jd = 1/10 = 10

Jd = 1/10 = .1

100

1

100

100
Master
1%

100 * .01 = 10 rows

Master
1%

100 * .01 = 10 rows

Join Filter ratios
FK

Detail

PK
Master

A

Jm = 1

Jm = “Join master ratio”
Jm = 1
when PK/FK relationship
Jm < 1
possible when not PKFK relationship

VST Steps Summary
1.Diagram tables
2.Draw connectors for each join
3.Calculate filter ratio

OD

S

P

F O

ODT

30%

A

F C

OT

0.02%

Child
Execution Path
• Start at the most selective join filter
• Join to keep the running result set size small

Child

Parent

Parent

VST Steps Summary
Many to Many relationships = problems

If many to many then
calculate two table join sizes

11/25/2013

VST steps summary
Note
• Don’t talk about table sizes
– Talk about filter ratios
– Two table join sizes
– Possibly look at join ratios

• Don’t talk about index existence
– Finding optimal path indicates indexes needed

Step 2: Get Explain Plan

Trace fie

-------------------------------------------------------------------------------------------| Id | Operation
| Name
| Starts | E-Rows | A-Rows |
--------------------------------------------------------------------------------------------|
1 | HASH GROUP BY
|
|
1 |
1 |
1 |
|* 2 |
FILTER
|
|
1 |
|
1909 |
|* 3 |
TABLE ACCESS BY INDEX ROWID
| PS_RETROPAYPGM_TBL |
1 |
1 |
3413 |
|
4 |
NESTED LOOPS
|
|
1 |
165 |
6827 |
|* 5 |
HASH JOIN
|
|
1 |
165 |
3413 |
|* 6 |
HASH JOIN
|
|
1 |
165 |
3624 |
|
7 |
TABLE ACCESS BY INDEX ROWID | WB_JOB
|
1 |
242 |
2895 |
|
8 |
NESTED LOOPS
|
|
1 |
233 |
2897 |
|
9 |
TABLE ACCESS BY INDEX ROWID| PS_PAY_CALENDAR
|
1 |
1 |
1 |
|* 10 |
INDEX RANGE SCAN
| PS0PAY_CALENDAR
|
1 |
1 |
1 |
|* 11 |
INDEX RANGE SCAN
| WBBJOB_B
|
1 |
286 |
2895 |
|* 12 |
TABLE ACCESS FULL
| WB_RETROPAY_EARNS |
1 | 27456 |
122K|
| 13 |
TABLE ACCESS FULL
| PS_RETROPAY_RQST
|
1 | 13679 | 13679 |
|* 14 |
INDEX RANGE SCAN
| PS#RETROPAYPGM_TBL |
3413 |
1 |
3413 |
| 15 |
SORT AGGREGATE
|
|
1791 |
1 |
1791 |
| 16 |
FIRST ROW
|
|
1791 |
1 |
1579 |
|* 17 |
INDEX RANGE SCAN (MIN/MAX)
| WB_JOB_F
|
1791 |
1 |
1579 |
| 18 |
SORT AGGREGATE
|
|
1539 |
1 |
1539 |
| 19 |
FIRST ROW
|
|
1539 |
1 |
1539 |
|* 20 |
INDEX RANGE SCAN (MIN/MAX) | WB_JOB_G
|
1539 |
1 |
1539 |
---------------------------------------------------------------------------------------------

VST

Predicate Information (identified by operation id):
---------------------------------------------------

2
3
5
6
10
11
12
14
17
20

-

filter(("B"."EFFDT"= AND "B"."EFFSEQ"=))
filter("E"."OFF_CYCLE"="A"."PAY_OFF_CYCLE_CAL")
access("D"."RETROPAY_SEQ_NO"="C"."RETROPAY_SEQ_NO")
access("C"."EMPLID"="B"."EMPLID" AND "C"."EMPL_RCD#"="B"."EMPL_RCD#")
access("A"."RUN_ID"='PD2' AND "A"."PAY_CONFIRM_RUN"='N')
access("B"."COMPANY"="A"."COMPANY" AND "B"."PAYGROUP"="A"."PAYGROUP")
filter(("C"."RETROPAY_PRCS_FLAG"='C' AND "C"."RETROPAY_LOAD_SW"='Y'))
access("E"."RETROPAY_PGM_ID"="D"."RETROPAY_PGM_ID")
access("F"."EMPLID"=:B1 AND "F"."EMPL_RCD#"=:B2 AND "F"."EFFDT"<=:B3)
access("G"."EMPLID"=:B1 AND "G"."EMPL_RCD#"=:B2 AND "G"."EFFDT"=:B3)

LOAD
Max CPU
(yard stick)
Top Activity

SQL

Sessions

Summary

1.Database - AAS

2.SQL - VST

10053
Table: A Alias: A
Best:: AccessPath: TableScan
Cost: 92.18 Degree: 1
Resp: 92.18 Card: 999.00

Best:: AccessPath: IndexRange
Index: B_V2
Resp: 4.00 Card: 1.00

Best:: AccessPath: IndexFFS
Index: C_PK_CON
Resp: 35.57 Card: 59999.00

A
$: 92
#: 999

B
$: 4
#: 1

C

B A C
B C A
C
C
A
A

A
B
C
B

B
A
B
C

Join order[1]: B[B]#0 A[A]#1 C[C]#2
***************
Now joining: A[A]#1
***************
Best:: JoinMethod: NestedLoop
Cost: 6.00 Degree: 1 Resp: 6.00 Card: 1.00 Bytes: 16
***************
Now joining: C[C]#2
***************
Best:: JoinMethod: NestedLoop
Cost: 6.00 Degree: 1 Resp: 6.00 Card: 1.00 Bytes: 21
***********************
Best so far: Table#: 0 cost: 4.0020 card: 1.0000 bytes: 12
Table#: 1 cost: 6.0031 card: 1.0000 bytes: 16
Table#: 2 cost: 6.0032 card: 1.0000 bytes: 21

Total: 14.0083

$: 5428
#: 59999

alter session set events='10053 trace name context forever';

No Unique Constraint

NL:$: 18
#: 1

NL:$: 18.01
#: 1

NL: $: 10
#: 1

C
$:

NL: $: 10.01
#: 1

A

8

Join order[6]: C[C]#2 A[A]#1 B[B]#0
Join order aborted: cost > best plan cost

$: 8

B

A

B

$: 4
#: 1

$: 6
B A C

$: 4
#: 1

C

C
$: 6.01
B C A

A

$: 5428 X
#: 59999
C A B

HJ: $: 10
#: 1

C
$: 5428 X
#: 59999
C B A

B

A
$: 75 X
#: 999
A C B

C

A
$: 75 X
#: 999
A BC

B

Unique Constraint

NL:$: 16
#: 1

NL:$: 18
#: 1
NL: $: 10
#: 1

C
$:

NL: $: 10
#: 1

A

8

Join order[6]: C[C]#2 A[A]#1 B[B]#0
Join order aborted: cost > best plan cost

$: 7

B

A

B

$: 4
#: 1

$: 6
B A C

$: 4
#: 1

C

C
$: 5

A

$: 5428 X
#: 59999
C A B

B C A

HJ: $: 10
#: 1

C
$: 5428 X
#: 59999
C B A

B

A
$: 75 X
#: 999
A C B

C

A
$: 75 X
#: 999
A BC

B

Types of Subqueries
• subquery
– query in where clause
– Select 1 from dual where 1 = (select 1 from dual);

• correlated subquery
– Subquery uses fields from outer query
– Select 1 from dual a where 1=(select 1 from dual b where
a.dummy=b.dummy)

• scalar subquery
– Query in select list
– Returns 1 value or null , ie scalar
– Select (select 1 from dual) from dual;

• inline views
– Query in from clause
– Select 1 from dual a, (select 1 from dual ) b;

• Layout tables Graphically
– Details tables above Masters
– One to one relations side by side
– Many to many any which way

Many to
single value

Join Set Sizes
Join

type

max result set size
possible

notes

one-to-scalar

A

this is equivalent to a filter on A
table B returns one value like a max(), min(),
count() etc

one-to-one

min(rA,rB)

one-to-many

rA
Joining from A to B will not increase the result set
size

many-to-many

rA*rB
------------------------min(ndv(A),ndv(B))

The more duplicates in both tables the greater
chance
the result set size will explode

We can say a lot about the join set size between two tables just by the
join type and the number of rows in the table and NDV in the join column,
but the easiest way most often is just to extract the two table joins and
run a count(*) on that subset of the query
129

11/25/2013

Join Filter ratios
FK

detail

A

Jd ~> 1

Jd ~> 1

B Jm <= 1

B Jm=1
PK

A

Master
Unique
index or
unique
constraint

Jd,
Jm
Jm
Jm

= join detail ratio , the avg number of rows returned when joining from the master to child
= master join ratio, the avg number of rows returned joining from the child into the master.
= 1 when PK/FK relations,
< 1 possible when unique indexes or unique constraints instead of PK/FK.

Jd < 1 rare but acts as filter instead of a multiplier

Hash Join vs NL
NL
X
Y

HJ
X
Y

Table X Driving Table (smaller set)
Col A
Col B
NL

Table X

Filter Index
Join Index

Col A
Col B

Table Y
Col C
Col D

HJ
Filter &
Join Index

Build Hash Table

Filter Index
Join Index

Drive from Table Y off of set of rows
returned from filter on column C
Nested loops into Table X on Index on
Join
Filter results without Index even though
index on filter column

Probe Hash Table

Filter Index
Join Index

Table Y

Col C
Col D

Filter Index
Join Index

Create hash result set on Table Y from
filter on column C
Probe hash result set with rows from
filter on table X on column A

Left Deep HJ
HJ

HJ
HJ

T1

T4
T3

T2

HJ
HJ
HJ
T1
T2
T3
T4

T4

Hashed probe
Hashed probe

T1
Hashed

probe

T3
T2

Right Deep HJ
HJ

HJ
T4
HJ
T3
HJ
T2
T1

HJ

T4

HJ

T3
T2

T1

T4

probe

hashed
T3

probe

probe

hashed
T2
hashed

T1

Nest Loops left deep

NL
probed
NL
probed
NL

T3

probed
T1

T2

T4

NL
NL
NL
T1
T2
T3
T4

Nest Loops can’t be right deep
NL
probed
T1

NL
probed
T2

NL
probed
T3

T4

NL
T1
NL
T2
NL
T3
T4

Bushy
select t3id, a.*, b.* from
( select /*+ no_merge */
t2.id t2id, t1.data t1data, t2.data t2data
from t1, t2
where t1.id = t2.id ) a,
( select /*+ no_merge */
t3.id t3id, t3.data t3data, t4.data t4data
from t3, t4
where t3.id = t4.id ) b
where a.t2id=b.t3id
/

HJ
NL
T1

NL

T2

T3

T4

HJ
NL
T1
T2
NL
T3
T4

DOAG: Visual SQL Tuning

More Related Content

What's hot

Viewers also liked

Similar to DOAG: Visual SQL Tuning

More from Kyle Hailey

Recently uploaded

DOAG: Visual SQL Tuning

Editor's Notes