Understanding index

Understanding BTree Index

主講人:申建忠

About me

frank_shen@uuu.com.tw 申建忠(Frank)

Agenda
• What is INDEX?
• BTree/Bitmap INDEX
• Why Optimizer does not use my INDEX?
• Constraint and INDEX
• How to reorganize INDEX
• TWO Myths about INDEX
• PCTFREE limit
• Reorganization is MUST

What are not include

• Bitmap Index internal
• Index-Organized Table
• Partitioned Index
• Function Based Index
• Bitmap Join Index

What is INDEX
• An index is an optional structure,
associated with a table or table cluster,
that can sometimes speed data access. By
creating an index on one or more columns of
a table, you gain the ability in some cases
to retrieve a small set of randomly
distributed rows from the table. Indexes
are one of many means of reducing disk I/O.
Oracle® Database Concepts 11g Release 2 (11.2)
-3 Indexes and Index-Organized Tables

BTree Structure
• BTree:Balanced Tree(B+Tree)
• Balance:
• The distance from root to any leaf node is always the
same.
• Height:The distance from root to leaf(max:24)
• Difference between BTree and Bitmap Index
• Index Entry Structure.
• Branch:
• Storage necessary information that can pointer to
appropriate leaf.
• Root:The top-level branch.
• Leaf:Storage index entries.

BTree Structure
Root Top-level branch

Branch Level
Height

Branch

Branch

Leaf
Index Entry Index Entry Index Entry Index Entry Index Entry Index Entry Index Entry Index Entry Index Entry Index Entry Index Entry Index Entry

smallest biggest

SQL> CREATE TABLE frank.t1
(a NUMBER,b NUMBER,c Date);
BTree Non-Unique Index

SQL> CREATE INDEX frank.t1_btree_idx ON hr.t1(a);
Bitmap Non-Unique Index

SQL> CREATE BITMAP INDEX frank.t1_bitmap_idx
ON hr.t1(b);

SQL> INSERT INTO frank.t1 VALUES(1,1,SYSDATE);
SQL> COMMIT;

SQL> SELECT object_name,object_id
FROM dba_objects
WHERE object_type=‘INDEX’ AND owner=‘FRANK’
AND object_name IN (T1_BTREE_IDX’,‘T1_BITMAP_IDX’);

OBJECT_NAME

OBJECT_ID
------------------------------ ----------------
T1_BTREE_IDX

76878
T1_BITMAP_IDX

76879 Display Index BTree Structure

SQL> ALTER SESSION SET EVENTS
‘IMMEDIATE TRACE NAME TREEDUMP LEVEL 76878’;

leaf: 0x1000253 16777811 (0: nrow: 3 rrow: 3) --BTree Index

‘IMMEDIATE TRACE NAME TREEDUMP LEVEL 76879’;

leaf: 0x100025b 16777819 (0: nrow: 2 rrow: 2) --Bitmap Index

Where is my Trace ﬁle?
SQL> SHOW PARAMETER user_dump_dest

NAME TYPE VALUE
---------------------- ------- -----------------------------------------------------
user_dump_dest string /u01/app/oracle/diag/rdbms/orcl/orcl/trace

SQL> SELECT spid FROM v$process
WHERE addr=(SELECT paddr FROM v$session
WHERE sid=(SELECT sid FROM v$mystat
WHERE rownum<=1));

SPID
----------

3565

--user traceﬁle named format : {SID}_ora_{ProcessID}.trc
/u01/app/oracle/diag/rdbms/orcl/orcl/trace/orcl_ora_3565.trc

SQL> CREATE OR REPLACE FUNCTION
get_block_address(p_dba NUMBER) RETURN VARCHAR2
2 IS
3 v_file NUMBER;
Convert Data Block Addrss(Decimal)
4 v_block NUMBER; to File number and Block number
5 BEGIN
6 v_file := DBMS_UTILITY.DATA_BLOCK_ADDRESS_FILE(p_dba);
7 v_block := DBMS_UTILITY.DATA_BLOCK_ADDRESS_ BLOCK(p_dba);
8 RETURN 'file: '||v_file||','||'block: '||v_block;
9 END;
10 /

Function created.

SQL> SELECT get_block_address(25166340) FROM dual;
GET_BLOCK_ADDRESS(25166340)
--------------------------------------------------------------------------------
Force DBWR write down
file: 6,block: 516
Dirty buffers to datafile
Dump index block content
SQL> ALTER SYSTEM CHECKPOINT;
SQL> ALTER SYSTEM DUMP DATAFILE 6 BLOCK 516;

BTree Index Dump
col 0 col 1
Header Key Value Rowid

row#0[8020] ﬂag: ------, lock: 2, len=12 Header
col 0; len 2; (2): c1 02 Key Value : 1
col 1; len 6; (6): 01 00 02 4f 00 00 Rowid
col 1; len 6; (6): 01 00 02 4f 00 02 Rowid
col 1; len 6; (6): 01 00 02 4f 00 01 Rowid

Bitmap Index Dump
col 0 col 1 col 2 col 3
Header Key Value Start Rowid End Rowid Bitmap Segs

col 1; len 6; (6): 01 00 02 4f 00 00 Start Rowid
col 2; len 6; (6): 01 00 02 4f 00 07 End Rowid
col 3; len 2; (2): c8 05 Encoded Bitmap Segment
row#1[7989] ﬂag: ------, lock: 0, len=21
col 1; len 6; (6): 01 00 02 4f 00 00
col 2; len 6; (6): 01 00 02 4f 00 07
col 3; len 1; (1): 01

DML on TABLE
• Manual DML on TABLE
• Automatic maintenance on INDEX
• INSERT on TABLE
• INSERT on INDEX
• DELETE on TABLE
• DELETE on INDEX
• UPDATE indexed column value on TABLE
• DELETE then INSERT on INDEX

INSERT rows(BTree)
(a NUMBER,b CHAR(200),c DATE);
SQL> CREATE INDEX frank.t2_idx ON frank.t2(a);

leaf: 0x1800204 25166340 (0: nrow: 0 rrow: 0)
leaf(root)

1.empty leaf block
2.only leaf block,no branch need

Insert ﬁrst row
SQL> insert into frank.t2 values(1,1,sysdate);

leaf: 0x1800204 25166340 (0: nrow: 1 rrow: 1)
leaf(root)

1 AAASxDAAEAAAAIPAAA

Insert another rows

leaf: 0x1800204 25166340 (0: nrow: 2 rrow: 2)
leaf(root)

2 AAASxDAAEAAAAIPAAB


Leaf block full

leaf block full

leaf(25166340) 4 AAASxDAAEAAAAIPAAD

3 AAASxDAAEAAAAIPAAC



Index Growing - 1
branch: 0x1800204 25166340 (0: nrow: 2, level: 1)
leaf: 0x1800218 25166360 (-1: nrow: 3 rrow: 3)
leaf: 0x1800207 25166343 (0: nrow: 1 rrow: 1)
Root block is always branch(25166340)
on the same block
kbx Height:2
brlmc 4 BLevel:1
add new leaf blocks
3 AAASxDAAEAAAAIPAAC


1 AAASxDAAEAAAAIPAAA 4 AAASxDAAEAAAAIPAAD

leaf(25166360) leaf(25166343)

Continue Insert
root(25166340)
kbx
brlmc 4 7
add new leaf block
and leaf node splits
3 AAASxDAAEAAAAIPAAC 6 AAASxDAAEAAAAIPAAF

2 AAASxDAAEAAAAIPAAB 5 AAASxDAAEAAAAIPAAE 8 AAASxDAAEAAAAIPAAH

1 AAASxDAAEAAAAIPAAA 4 AAASxDAAEAAAAIPAAD 7 AAASxDAAEAAAAIPAAG

leaf(25166360) leaf(25166355) leaf(25166343)

Index Growing - 2

root(25166340)
kbx Root and leaf
brlmc 4 7 are both full

10 AAASxDAAEAAAAIPAAJ

3 AAASxDAAEAAAAIPAAC 6 AAASxDAAEAAAAIPAAF 9 AAASxDAAEAAAAIPAAI


1 AAASxDAAEAAAAIPAAA 4 AAASxDAAEAAAAIPAAD 7 AAASxDAAEAAAAIPAAG

leaf(25166360) leaf(25166355) leaf(25166343)

Index Growing - 3
root(25166340)
kbx
Height:3
brlmc 7 BLevel:2
add new branch blocks
kbx
kbx
brlmc 4 brlmc 10
branch(25166344) branch(25166345)



1 AAASxDAAEAAAAIPAAA 4 AAASxDAAEAAAAIPAAD 7 AAASxDAAEAAAAIPAAG 10 AAASxDAAEAAAAIPAAJ

leaf(25166360) leaf(25166355) leaf(25166343) leaf(25166356)
leaf block split

Leaf Node Splits
SQL> SELECT m.*,s.name
FROM v$mystat m,v$statname s
WHERE m.statistic#=s.statistic# AND s.name LIKE '%leaf%';

SID STATISTIC#

VALUE NAME
---------- ---------------- ----------- --------------------------

1

412

0 leaf node splits

1

413

0 leaf node 90-10 splits

root

1 Rowid 11 Rowid
3 Rowid 13 Rowid
21 Rowid
5 Rowid 15 Rowid
7 Rowid 17 Rowid 21 Rowid

root
90-10 split

1 Rowid 11 Rowid
root
3 Rowid 13 Rowid
5 Rowid 15 Rowid
7 Rowid 17 Rowid

1 Rowid
3 Rowid 12 Rowid
12 Rowid

50-50 split

DELETE rows
SQL> DELETE frank.t2 WHERE a=5;
root(25166340)
kbx
brlmc 7

kbx
kbx
brlmc 4 brlmc 10
branch(25166344) branch(25166345)




UPDATE indexed columns
SQL> UPDATE frank.t2 SET a=11 WHERE a=1;
root(25166340)
kbx
brlmc 7

kbx
kbx
brlmc 4 brlmc 10
branch(25166344) branch(25166345)

3 AAASxDAAEAAAAIPAAC 6 AAASxDAAEAAAAIPAAF 9 AAASxDAAEAAAAIPAAI 2.insert new entry
2 AAASxDAAEAAAAIPAAB 5 AAASxDAAEAAAAIPAAE 8 AAASxDAAEAAAAIPAAH 11 AAASxDAAEAAAAIPAAA

1.delete old entry

Why Optimizer don’t
use my index ?
• just COST
• Index unusable
• index invisible
• Index unavailable

Optimizer Cost
According to the CPU costing model:
Cost = ( #SRds * sreadtim +
#MRds * mreadtim +
#CPUCycles / cpuspeed ) / sreadtim

where:
• #SRDs is the number of single block reads
• #MRDs is the number of multi block reads
• #CPUCycles is the number of CPU Cycles *)
• sreadtim is the single block read time
• mreadtim is the multi block read time
• cpuspeed is the CPU cycles per second
CPUCycles includes CPU cost of query processing (pure CPU cost) and CPU cost of
data retrieval (CPU cost of the buffer cache get).
This model is straightforward for serial execution.
Oracle9i Database Performance Tuning Guide and
Reference Release 2 (9.2) - 9 Using EXPLAIN PLAN

SQL> CREATE TBALE frank.t3 AS
SELECT LEVEL a,’b’||LEVEL b,SYSDATE c
FROM dual
CONNECT BY LEVEL<=500000;

SQL> CREATE INDEX frank.t3_a_idx ON frank.t3(a);

SQL> EXECUTE dbms_stats.gather_table_stats(‘FRANK’,’T3’);

SQL> SELECT COUNT(*) FROM frank.t3;

COUNT(*)
-----------------
500000

SQL> SELECT num_rows,num_blocks
2 FROM dba_tables
3 WHERE owner=‘FRANK’ AND table_name=’T3’;

NUM_ROWS NUM_BLOCKS
-------------------- --------------------
500000 1820

index scan root

branch branch

1 rowid 4 rowid 7 rowid 10 rowid 13 rowid 16 rowid 19 rowid 22 rowid

1 a 4 d 7 g 10 j
2 b 5 e 8 h 11 k
full table scan 3 c 6 f 9 i 12 l

13 m 16 p 19 s 22 v
14 n 17 q 20 t 23 w
15 o 18 r 21 u 24 x

‘10053 TRACE NAME CONTEXT FOREVER,LEVEL 8’;

SQL> SELECT a,b,c FROM frank.t3 WHERE a=1;

1 rows selected.
---------------------------------------------------------------------+----------------------------------------+
| Id | Operation | Name | Rows | Bytes | Cost | Time |
---------------------------------------------------------------------+----------------------------------------+
| 0 | SELECT STATEMENT | | | | 4| |
| 1 | TABLE ACCESS BY INDEX ROWID | T3 | 1 | 21 | 4 | 00:00:01|
| 2 | INDEX RANGE SCAN | T3_A_IDX| 1 | | 3 | 00:00:01|
----------------------------------------------------------------------+----------------------------------------+
Predicate Information:
-----------------------------
2 - access("A"=1)

SINGLE TABLE ACCESS PATH
Single Table Cardinality Estimation for T3[T3]
Column (#1): A(
AvgLen: 5 NDV: 500000 Nulls: 0 Density: 0.000002 Min: 1 Max: 500000
Table: T3 Alias: T3
Card: Original: 500000.000000 Rounded: 1 Computed: 1.00 Non Adjusted: 1.00
Access Path: TableScan
Cost: 500.47 Resp: 500.47 Degree: 0
Cost_io: 495.00 Cost_cpu: 112961061
Resp_io: 495.00 Resp_cpu: 112961061
Access Path: index (AllEqRange)
Index: T3_A_IDX
resc_io: 4.00 resc_cpu: 28876
ix_sel: 0.000002 ix_sel_with_ﬁlters: 0.000002
Best:: AccessPath: IndexRange
Index: T3_A_IDX
Cost: 4.00 Degree: 1 Resp: 4.00 Card: 1.00 Bytes: 0


SQL> SELECT a,b,c FROM frank.t3 WHERE a>1;

499999 rows selected.
--------------------------------------------------+-----------------------------------------+
--------------------------------------------------+-----------------------------------------+
| 0 | SELECT STATEMENT | | | | 501 | |
| 1 | TABLE ACCESS FULL | T3 | 488K | 10M | 501 | 00:00:07 |
--------------------------------------------------+------------------------------------------+
----------------------
1 - ﬁlter("A">1)

Column (#1): A(
Table: T3 Alias: T3
Card: Original: 500000.000000 Rounded: 499999 Computed: 499999.00 Non
Adjusted: 499999.00
Cost_io: 495.00 Cost_cpu: 132960981
Resp_io: 495.00 Resp_cpu: 132960981
Access Path: index (RangeScan)
Index: T3_A_IDX
resc_io: 2897.00 resc_cpu: 215630422
Cost: 2907.45 Resp: 2907.45 Degree: 1
Best:: AccessPath: TableScan


SQL> SELECT a,b,c FROM frank.t3 WHERE a>490000;

10000 rows selected.
---------------------------------------------------------------------+----------------------------------------+
---------------------------------------------------------------------+----------------------------------------+
| 1 | TABLE ACCESS BY INDEX ROWID | T3 | 10K | 205K| 61 | 00:00:01|
| 2 | INDEX RANGE SCAN | T3_A_IDX| 10K | | 25 | 00:00:01|
----------------------------------------------------------------------+----------------------------------------+
-----------------------------
2 - access("A">490000)

Column (#1): A(
Table: T3 Alias: T3
Card: Original: 500000.000000 Rounded: 10000 Computed: 10000.02 Non
Adjusted: 10000.02
Cost_io: 495.00 Cost_cpu: 113361021
Resp_io: 495.00 Resp_cpu: 113361021
Access Path: index (RangeScan)
Index: T3_A_IDX
resc_io: 61.00 resc_cpu: 4334798
Best:: AccessPath: IndexRange
Index: T3_A_IDX

Index properties
• Usability
Indexes are usable (default) or unusable. An unusable index is not
maintained by DML operations and is ignored by the optimizer. An
unusable index can improve the performance of bulk loads. Instead of
dropping an index and later re-creating it, you can make the index
unusable and then rebuild it. Unusable indexes and index partitions
do not consume space. When you make a usable index unusable, the
database drops its index segment.
• Visibility
Indexes are visible (default) or invisible. An invisible index is
maintained by DML operations and is not used by default by the
optimizer. Making an index invisible is an alternative to making it
unusable or dropping it. Invisible indexes are especially useful for
testing the removal of an index before dropping it or using indexes
temporarily without affecting the overall application.
Oracle® Database Concepts 11g Release 2 (11.2)
-3 Indexes and Index-Organized Tables

Unusable Index
SQL> SELECT status
FROM dba_indexes
WHERE owner=‘FRANK’ and table_name=’T3’;

STATUS
---------------
UNUSABLE


1 rows selected.
--------------------------------------------------+-----------------------------------------+
--------------------------------------------------+-----------------------------------------+
| 1 | TABLE ACCESS FULL | T3 | 1| 21 | 500 | 00:00:06 |
--------------------------------------------------+-----------------------------------------+
----------------------
1 - ﬁlter("A"=1)

Don’t skip unusable Index
SQL> SHOW PARAMETER skip_unusable_indexe

NAME TYPE VALUE
------------------------------------ ----------- ------------------------------
skip_unusable_indexes boolean TRUE

SQL> ALTER SYSTEM SET skip_unusable_indexes=FALSE;

Session altered.

SELECT a,b,c FROM frank.t1 WHERE a=1;
*
ERROR at line 1:
ORA-01502: index 'FRANK.T3_A_IDX' or partition of such index is
in unusable state

Invisible Index
SQL> ALTER INDEX frank.t3_a_idx INVISIBLE;

SQL> SELECT status,visibility FROM dba_indexes
2 WHERE owner='FRANK' AND table_name='T3';

STATUS VISIBILIT
----------- -------------
VALID INVISIBLE


1 rows selected.
--------------------------------------------------+-----------------------------------------+
--------------------------------------------------+-----------------------------------------+
| 1 | TABLE ACCESS FULL | T3 | 1| 21 | 500 | 00:00:06 |
--------------------------------------------------+-----------------------------------------+
----------------------
1 - ﬁlter("A"=1)

SQL> SHOW PARAMETER optimizer_use_invisible_indexes

NAME TYPE VALUE
------------------------------------------- ----------- ------------------------------
optimizer_use_invisible_indexes boolean FLASE

SQL> ALTER SESSION SET optimizer_use_invisible_indexes=TRUE;


1 rows selected.
---------------------------------------------------------------------+----------------------------------------+
---------------------------------------------------------------------+----------------------------------------+
| 0 | SELECT STATEMENT | | | | 4| |
| 1 | TABLE ACCESS BY INDEX ROWID| T3 | 1 | 21 | 4 | 00:00:01 |
| 2 | INDEX RANGE SCAN | T3_A_IDX| 1 | | 3 | 00:00:01 |
----------------------------------------------------------------------+----------------------------------------+
-----------------------------
2 - access("A"=1)

SQL> ALTER INDEX frank.t3_a_idx VISIBLE;

Index unavailable
• 條件式為IS NULL或IS NOT NULL或NOT
• SELECT a,b,c FROM frank.t1 WHERE a IS NULL;

• SELECT a,b,c FROM frank.t1 WHERE a != 1;

• indexed column被包含在運算式或函數
• SELECT a,b,c FROM frank.t1 WHERE a+100=101;

• SELECT a,b,c FROM frank.t1 WHERE TO_CHAR(a)=’1’;

• LIKE條件式為’%xx’或’%xx%’
• SELECT a,b,c FROM frank.t1 WHERE b LIKE ‘%12’;

Implicit type conversion
SQL> SELECT a,b,c FROM frank.t3 WHERE a = '100000';

AB C
------------- ----------------------------------------- ------------
100000 b100000 07-FEB-12

---------------------------------------------------------------------+----------------------------------------+
---------------------------------------------------------------------+----------------------------------------+
| 1 | TABLE ACCESS BY INDEX ROWID | T3 | 10K | 205K| 61 | 00:00:01|
| 2 | INDEX RANGE SCAN | T3_A_IDX| 10K | | 25 | 00:00:01|
----------------------------------------------------------------------+----------------------------------------+
Predicate Information: Happy Ending
-----------------------------
2 - access("A"=100000)

Implicit type conversion
SQL> SELECT a,b,c FROM frank.t3 WHERE a LIKE '100000';

AB C
------------- ----------------------------------------- ------------
100000 b100000 07-FEB-12

--------------------------------------------------+-----------------------------------------+
--------------------------------------------------+-----------------------------------------+
| 1 | TABLE ACCESS FULL | T3 | 1 | 21 | 500 | 00:00:06 |
--------------------------------------------------+-----------------------------------------+
Predicate Information: Bad Ending
----------------------
1 - ﬁlter(TO_CHAR(“A”)=‘100000’)

Index and Constraints
• Primary Key and Unique Constraints
• Index must exist
• Improve integrity constraint check time
• Foreign Key Constraints
• suggest to create btree index on FK
• Improve integrity constraint check time
• Less Lock Contentions on Child Table
• Prevent Dead Lock

Enable Primary Key/
Unique Constraint

Enable PK/UK
Is index Yes Using existing
existed on pk/uk
Constraint index
columns?

No

Create Unique No Is constraint Yes Create Non-
Index deferrable? Unique Index

Default type:
SQL> CREATE TABLE frank.t4 NOT DEFERRABLE
(a NUMBER CONSTRAINT t4_a_pk PRIMARY KEY,
b VARCHAR2(10) CONSTRAINT t4_b_uk UNIQUE DEFERRABLE);

SQL> SELECT constraint_name,constraint_type,deferrable,index_name
FROM dba_constraints
WHERE owner='HR' and table_name='T3';

CONSTRAINT_NAME C DEFERRABLE INDEX_NAME
------------------------------ - -------------------------- --------------------
T4_B_UK

U DEFERRABLE

T4_B_UK
T4_A_PK

P NOT DEFERRABLE T4_A_PK

SQL> SELECT index_name,index_type,uniqueness
FROM dba_indexes

INDEX_NAME

INDEX_TYPE

UNIQUENES
------------------------------ --------------------------- ------------------
T4_A_PK

NORMAL

UNIQUE
T4_B_UK

NORMAL

NONUNIQUE

Disable/Drop Primary
Key/Unique Constraint

Disable/Drop
PK/UK
Is Index Yes Is Index Yes Drop exising
auto-created? uniqueness? Index
Constraint

No No
Keep existing Keep existing
Index Index

SQL> ALTER TABLE frank.t4 DROP CONSTRAINT t4_a_pk; --Drop PK

SQL> ALTER TABLE frank.t4 DROP CONSTRAINT t4_b_uk; --Drop UK

WHERE owner='FRANK' and table_name='T4';

no rows selected

FROM dba_indexes

INDEX_NAME

INDEX_TYPE

UNIQUENES
------------------------------ --------------------------- ------------------
T4_B_UK

NORMAL

NONUNIQUE
After drop/disable PK/UK
constraint, auto-created unique
index will be dropped.

SQL> CREATE UNIQUE INDEX frank.t4_a_idx ON frank.t4(a); --Create index

SQL> ALTER TABLE frank.t4 ADD CONSTRAINT t4_a_pk PRIMARY KEY(a);

WHERE owner='FRANK' and table_name='T4';

CONSTRAINT_NAME

DEFERRABLE C INDEX_NAME
------------------------------ - -------------------------- --------------------
T4_A_PK

P NOT DEFERRABLE T4_A_IDX --PK use existing index

SQL> ALTER TABLE frank.t4 DROP CONSTRAINT t4_a_pk; --Drop PK constraint

FROM dba_indexes

INDEX_NAME

INDEX_TYPE

UNIQUENES Manual created
------------------------------ --------------------------- ------------------ index would not
T4_A_IDX

NORMAL

UNIQUE be autodropped
T4_B_UK

NORMAL

NONUNIQUE

No Index On FK cols
INSERT
DML
DETELE
UPDATE FK value
3.TX:X
2.TM:RX
1.TM:RX
PK 1 a1
Parent 1 a2
2 b 2 b1
1 a FK Child

Insert into Parent Table
INSERT
DML
DETELE
UPDATE FK value
3.TX:X
INSERT
2.TM:RX 2.TM:RX
1.TM:RX
PK 1 a1
3 c 1 a2
1.TM:RX
2 b 2 b1
3.TX:X
1 a FK

Update Parent Table
INSERT
DML
DETELE
UPDATE FK value
2.TM:RX
UPDATE
2.TM:S(release after hold)
1.TM:RX
PK 1 a1
3=>4 c 1 a2
1.TM:RX
2 b 2 b1
3.TX:X
1 a FK

Delete Parent Table - 1
INSERT
DML
DETELE
UPDATE FK value
2.TM:RX
DELETE
2.TM:S(release after hold)
1.TM:RX
PK 1 a1
3 c 1 a2
1.TM:RX
2 b 2 b1
3.TX:X
1 a FK

No Action

INSERT
DML
DETELE
UPDATE FK value
2.TM:RX
DELETE
2.TM:SRX -> RX
1.TM:RX
PK 1 a1
3 c 1 a2
1.TM:RX
2 b 2 b1
1 a FK
On Delete Cascade

INSERT
DML
DETELE
UPDATE FK value
3.TX:X
DELETE
2.TM:RX 2.TM:RX
1.TM:RX
PK 1 a1
3 c 1 a2
1.TM:RX
2 b 2 b1
3.TX:X
1 a FK
On Delete Cascade

FK without Index
SQL> INSERT INTO child VALUES(2,’b2’);
1 row created.
SQL> DECLARE
PRAGMA autonomous_transaction;
BEGIN
DELETE parent WHERE a=3;
COMMIT;
END;
--after wait a short time
DECLARE
*
ERROR at line 1:
ORA-00060: deadlock detected while waiting for
resource
ORA-06512: at line 4

Index On FK cols
BTree index
INSERT
DML
DETELE
UPDATE FK value
3.TX:X
2.TM:RX
1.TM:RX
PK 1 a1
Parent 1 a2
2 b 2 b1
1 a FK Child

Insert into Parent Table - 10g
BTree index
INSERT
DML
DETELE
UPDATE FK value
2.TX:X
INSERT
2.TM:RS 1.TM:RX
1.TM:RX
PK 1 a1
3 c 1 a2
1.TM:RX
2 b 2 b1
3.TX:X
1 a FK

Insert into Parent Table - 11g
BTree index
INSERT
DML
DETELE
UPDATE FK value
2.TX:X
INSERT
2.TM:RX 1.TM:RX
1.TM:RX
PK 1 a1
3 c 1 a2
1.TM:RX
2 b 2 b1
3.TX:X
1 a FK

Update Parent Table - 10g
BTree index
INSERT
DML
DETELE
UPDATE FK value
2.TX:X
UPDATE
2.TM:RS 1.TM:RX
1.TM:RX
PK 1 a1
3=>4 c 1 a2
1.TM:RX
2 b 2 b1
3.TX:X
1 a FK

Update Parent Table - 11g
BTree index
INSERT
DML
DETELE
UPDATE FK value
2.TX:X
UPDATE
2.TM:RX 1.TM:RX
1.TM:RX
PK 1 a1
3=>4 c 1 a2
1.TM:RX
2 b 2 b1
3.TX:X
1 a FK

Delete Parent Table - 10g
BTree index
INSERT
DML
DETELE
UPDATE FK value
2.TX:X
DELETE
2.TM:RS 1.TM:RX
1.TM:RX
PK 1 a1
3 c 1 a2
1.TM:RX
2 b 2 b1
3.TX:X
1 a FK

Delete Parent Table - 11g
BTree index
INSERT
DML
DETELE
UPDATE FK value
2.TX:X
DELETE
2.TM:RX 1.TM:RX
1.TM:RX
PK 1 a1
3 c 1 a2
1.TM:RX
2 b 2 b1
3.TX:X
1 a FK

FK with Index
SQL> INSERT INTO child VALUES(2,’b2’);

1 row created.

SQL> DECLARE
PRAGMA autonomous_transaction;
BEGIN
DELETE parent WHERE a=3;
COMMIT;
No more Deadlock
END;

PL/SQL procedure successfully completed.

Index Reorganization

• ALTER INDEX index_name REBUILD;
• ALTER INDEX index_name REBUILD
ONLINE;
• ALTER INDEX index_name COALESCE;
• DROP INDEX then CREATE INDEX

Rebuild or Coalesce?
Rebuild Index Coalesce Index

Quickly moves index to another tablespace Cannot move index to another tablespace

Higher cost: requires more disk space Lower costs: does not require more disk space

Creates new tree,shrinks height if applicable Coalesce leaf blocks within same branch of tree

Enables you to quickly change storage and
tablespace parameters without having to drop the Quickly frees up index leaf blocks for use
original index
Oracle® Database Administrator's Guide
11g Release 2 (11.2) - 21 Managing Indexes

Rebuild
Rebuild

SQL> ALTER INDEX xxx REBUILD;

DML TM:S

Rebuild online - 10g

e
lin
on
u ild
R eb

SQL> ALTER INDEX xxx REBUILD
ONLINE;

DML TM:S(release immediate)-Begin
TM:RS
TM:S(release immediate)-End

Rebuild online - 11g

e
lin
on
u ild
R eb

SQL> ALTER INDEX xxx REBUILD
ONLINE;

DML TX:S(ongoing transaction)-Begin
TM:RS
TX:S(ongoing transaction)-End

Height
Coalesce Index

SQL> ALTER INDEX xxx COALESCE;

TM:RS

DML

Index Related Myths
• PCTFREE always take effect
• pctfree only effect on creation
• ReOrganization is MUST
• Depends on deleted space can be reused
or not

INDEX Before INSERT
2 (a NUMBER,b CHAR(200),c DATE);

SQL> CREATE INDEX frank.t5_ab_idx ON frank.t5(a,b)
2 PCTFREE 50 --default:10
3 TABLESPACE ts2k;

SQL> INSERT INTO frank.t5
2 SELECT LEVEL,'b'||LEVEL,SYSDATE+LEVEL
3 FROM dual CONNECT BY LEVEL<=8;
SQL> COMMIT;

SQL> ANALYZE INDEX frank.t5_ab_idx VALIDATE STRUCTURE;

SQL> SELECT lf_rows,lf_blks,del_lf_rows,used_space,pct_used FROM index_stats;

LF_ROWS LF_BLKS DEL_LF_ROWS USED_SPACE PCT_USED
---------------- -------------- -------------------- ------------------- ----------------

8

1

0

1728

94

INDEX After INSERT
SQL> CREATE TABLE frank.t6 AS SELECT * FROM frank.t5;

SQL> CREATE INDEX frank.t6_ab_idx ON frank.t6(a,b)
2 PCTFREE 50
3 TABLESPACE ts2k;



---------------- -------------- -------------------- ------------------- ----------------

8

3

0

1748

24

branch: 0x1800224 25166372 (0: nrow: 3, level: 1) --TREEDUMP
leaf: 0x1800225 25166373 (-1: nrow: 3 rrow: 3)
leaf: 0x1800226 25166374 (0: nrow: 3 rrow: 3)
leaf: 0x1800227 25166375 (1: nrow: 2 rrow: 2)

INDEX maintenance
ignore PCTFEE setting
SQL> INSERT INTO frank.t6 VALUES(9,’b9’,SYSDATE);
SQL> INSERT INTO frank.t6 VALUES(10,‘b10’,SYSDATE);
SQL> INSERT INTO frank.t6 VALUES(11,‘b11’,SYSDATE); Index Entries
SQL> COMMIT;

leaf: 0x1800225 25166373 (-1: nrow: 3 rrow: 3)
leaf: 0x1800226 25166374 (0: nrow: 3 rrow: 3)
leaf: 0x1800227 25166375 (1: nrow: 8 rrow: 8)

2 (a NUMBER,b CHAR(200),c DATE);
SQL> CREATE INDEX frank.t7_ab_idx ON frank.t7(a,b) TABLESPACE ts2k;

SQL> INSERT INTO frank.t7
2 SELECT LEVEL,'b'||LEVEL,SYSDATE+LEVEL
3 FROM dual CONNECT BY LEVEL<=20;
SQL> COMMIT;



---------------- -------------- -------------------- ------------------ -----------------
20 3 0 4340 59

leaf: 0x1800254 25166420 (-1: nrow: 8 rrow: 8)
leaf: 0x180024e 25166414 (0: nrow: 8 rrow: 8)
leaf: 0x180024f 25166415 (1: nrow: 4 rrow: 4)

INDEX just be created
25166404
root

1,b1 Rowid 9,b9 Rowid
5,b5 Rowid 13,b13 Rowid 17,b17 Rowid
25166420 25166414 25166415

SQL> DELETE frank.t7 WHERE MOD(a,2)=0;

10 rows deleted.

SQL> COMMIT;



---------------- -------------- -------------------- ------------------ -----------------
20 3 10 4340 59

del_lf_rows/lf_rows=50%
If > 20% then reorganized index
Really need to do this ?

leaf: 0x1800254 25166420 (-1: nrow: 8 rrow: 4)
leaf: 0x180024e 25166414 (0: nrow: 8 rrow: 4)
leaf: 0x180024f 25166415 (1: nrow: 4 rrow: 2)

INDEX after Delete
25166404
root

25166420 25166414 25166415

SQL> INSERT INTO frank.t7 VALUES(4,’b4’,SYSDATE); index entries
SQL> COMMIT;


SQL> SELECT lf_rows,lf_blks,del_lf_rows,used_space,pct_used
2 FROM index_stats;

---------------- -------------- -------------------- ------------------ -----------------
20 3 6 4340 59

leaf: 0x1800254 25166420 (-1: nrow: 8 rrow: 8)
leaf: 0x180024e 25166414 (0: nrow: 8 rrow: 4)
leaf: 0x180024f 25166415 (1: nrow: 4 rrow: 2)

Reuse deleted space
25166404
root

Reuse deleted space
Reorganization is not necessary

25166420 25166414 25166415

SQL> INSERT INTO frank.t7 VALUES(31,‘b31’,SYSDATE); index entries
SQL> COMMIT;


SQL> SELECT lf_rows,lf_blks,del_lf_rows,used_space,pct_used
2 FROM index_stats;

---------------- -------------- -------------------- ------------------ -----------------
22 3 8 4772 65

leaf: 0x1800254 25166420 (-1: nrow: 8 rrow: 4)
leaf: 0x180024e 25166414 (0: nrow: 8 rrow: 4)
leaf: 0x180024f 25166415 (1: nrow: 6 rrow: 6)

Can’t reuse deleted space
25166404
root

Deleted space does not be reused
Reorganization is necessary

25166420 25166414 25166415

Q&A

• Http://www.uuu.com.tw
• Email:frank_shen@uuu.com.tw
• Blog://frankshenoracle.blogspot.com

Understanding index

More Related Content

What's hot (20)

Viewers also liked (8)

Similar to Understanding index (20)

More from Chien Chung Shen (11)

Recently uploaded (20)

Understanding index