Vous êtes sur la page 1sur 35

Physical Level of Databases:

B+-Trees

Adnan YAZICI Computer Engineering Department METU

DBMS , A. Yaz c

(Fall 2005)

1

B + -Tree Index Files

l

Disadvantage of indexed-sequential files:

performance degrades as file grows, since many overflow blocks get created. Periodic reorganization of entire file is required.

l

Advantage of B + -tree index files: automatically reorganizes itself with small and local changes, in the face of insertions and deletions. Reorganization of entire file is not required to maintain performance.

l

Disadvantage of B + -trees: extra insertion and deletion overhead, space overhead.

l

Advantages of B + -trees outweigh disadvantages, and they are used extensively.

DBMS , A. Yaz c

2

B + -Tree Index Files (Cont.)

l

All paths from root to leaf are of the same length

l

Each node that is not a root or a leaf has between [n/2] and n children.

l

Special cases:

– If the root is not a leaf, it has at least 2 children. – If the root is a leaf (that is, there are no other nodes in the tree), it can have between 0 and (n–1) values.

DBMS , A. Yaz c

3

B+-Trees as Dynamic Multi-level Indexes

B+-Tree is a variation of search trees that allow efficient insertion and deletion of new search values. •In B+-Tree , each node corresponds to a disk block. •Each node is kept between half-full and completely-full. •An insertion into a node that is not full is quite efficient; if a node is full the insertion causes a split into two nodes. •Splitting may propagate to other tree levels. •A deletion is efficient if a node does not become less than half full. •If a deletion causes a node to become less than half full, it must be merged with neighboring nodes.

DBMS , A. Yaz c

4

B + -Tree Node Structure

l Typical node

B + -Tree Node Structure l Typical node – K i are the search-key values –

– K i are the search-key values

– P i are pointers to children (for non-leaf nodes) or pointers to records or buckets of records (for leaf nodes).

l The search-keys in a node are ordered

K 1 < K 2 < K 3 <

DBMS , A. Yaz c

< K n1

5

Leaf Nodes in B + -Trees

l

For i = 1, 2,

., n, pointer P i either points to a file record with

search-key value K i , or to a bucket of pointers to file records, each record having search-key value K i . Only need bucket structure if search-key does not form a primary key.

l

If L i , L j are leaf nodes and i < j, L i ’s search-key values are less than L j ’s search-key values

l

P n points to next leaf node in search-key order

than L j ’s search-key values l P n points to next leaf node in search-key

DBMS , A. Yaz c

6

Non-Leaf Nodes in B + -Trees

l Non leaf nodes form a multi-level sparse index on the leaf nodes. For a non-leaf node with m pointers:

– All the search-keys in the subtree to which P 1 points are less than K 1

– For 2 i n – 1, all the search-keys in the subtree to which P i points have values greater than or equal to K i1 and less than K m1

P i points have values greater than or equal to K i – 1 and less

DBMS , A. Yaz c

7

Example of a B + -tree

Example of a B + -tree DBMS , A. Yaz c 8

DBMS , A. Yaz c

8

Example of B + -tree

Example of B + -tree ! " l All nodes must have between 2 and 4

! "

l

All nodes must have between 2 and 4 key values ( (n)/2 and n, with n = 5).

l

Root must have at least 2 children.

DBMS , A. Yaz c

9

Observations about B + -trees

l

Since the inter-node connections are done by pointers, “logically” close blocks need not be “physically” close.

l

The non-leaf levels of the B + -tree form a hierarchy of sparse indices.

l

The B + -tree contains a relatively small number of levels (logarithmic in the size of the main file), thus searches can be conducted efficiently.

l

Insertions and deletions to the main file can be handled efficiently, as the index can be restructured in logarithmic time (as we shall see).

DBMS , A. Yaz c

10

Queries on B + -Trees

l Find all records with a search-key value of k.

1. Start with the root node

1. Examine the node for the smallest search-key value > k.

2. If such a value exists, assume it is K j . Then follow P i to the child node

3. Otherwise k K m1 , where there are m pointers in the node. Then follow P m to the child node.

2. If the node reached by following the pointer above is not a leaf node, repeat step 1 on the node

3. Else we have reached a leaf node.

1. If for some i, key K i = k follow pointer P i to the desired record or bucket.

2. Else no record with search-key value k exists.

DBMS , A. Yaz c

11

Queries on B +- Trees (Cont.)

l In processing a query, a path is traversed in the tree from the root to some leaf node.

l If there are K search-key values in the file, the path is no longer than log n/2 (K) . l A node is generally the same size as a disk block, typically 4 kilobytes, and n is typically around 100 (40 bytes per index entry).

l

With 1 million search key values and n = 100, at most log 50 (1,000,000) = 4 nodes are accessed in a lookup.

l

Contrast this with a balanced binary free with 1 million search key values — around 20 nodes are accessed in a lookup – above difference is significant since every node access may need a disk I/O, costing around 20 milliseconds!

DBMS , A. Yaz c

12

B+ Tree

B+ Tree DBMS , A. Yaz c 13

DBMS , A. Yaz c

13

B+ Tree

l

B+ Tree Properties

l

B+ Tree Searching

l

B+ Tree Insertion

l

B+ Tree Deletion

DBMS , A. Yaz c

14

B+ Tree

– Balanced Tree

l

Same height for paths from root to leaf

l

Given a search-key K, nearly same access time for different K values

– B+ Tree is constructed by parameter n

l

Each Node (except root) has n/2 to n pointers

l

Each Node (except root) has n/2 -1 to n-1 search- key values

Case for n=3

K K 1 2 P 1 P 2 P 3
K
K
1
2
P 1
P 2
P 3

DBMS , A. Yaz c

P

General case for n

K K 1 K 2 n-1 1 P 2 P n P n-1
K
K 1
K 2
n-1
1
P 2
P n
P n-1

15

B+ Tree

l Search keys are sorted in order

– K 1 < K 2 < … <K n-1

•Non-leaf Node

–Each key-search values in subtree S i pointed by P i < K i , >=K i-1 Key values in S 1 < K 1 K 1 <= Key values in S 2 < K 2

K K
K
K

1

P 2

2

1 <= Key values in S 2 < K 2 K K 1 P 2 2

P 3

S 2
S 2
P 1 S 1
P 1
S 1
S 3
S 3

•Leaf Node

–P i points record or bucket with search key value K i –P n points to the neighbor leaf node

K K 1 2
K
K
1
2
i –P n points to the neighbor leaf node K K 1 2 P 1 P

P 1

P

3 …
3
P 2 Record of K 2 Record of K 2 …
P 2
Record of K 2
Record of K 2

Record of K 1

DBMS , A. Yaz c

16

B+ Tree: Search

l Given a search-value k

– Start from the root, look for the largest search-key value (K l ) in the node <= k

– Follow pointer P l+1 to next level, until reach a leaf node

K l <=k<K l+1 … K K K K 1 K 2 l l+1 n-1
K
l <=k<K l+1
K
K
K
K 1
K 2
l
l+1
n-1
P
2
P 3
P
P n
P n-1
l+1

P 1

– If k is found to be equal to K l in the leaf, follow P l to

search the record or bucket

DBMS , A. Yaz c

K l K l+1
K l
K l+1

P l

k = K l

Record of K l Record of K l …
Record of K l
Record of K l

17

B+ Tree:Insertion

l Overflow

– When the number of search-key values exceed n-1

–Leaf Node

•Split into two nodes:

7 9 13 15
7 9
13 15

Insert 8

–1 st node contains (n-1)/2 values –2 nd node contains remaining values –Copy the smallest search-key value of the 2 nd node to parent node

9 7 8 9 13 15
9
7 8
9
13 15

DBMS , A. Yaz c

18

B+ Tree: Insertion

l Overflow

– When number of search-key values exceed n-1

–Non-Leaf Node

•Split into two nodes:

7 9 13 15
7 9
13
15

Insert 8

–1 st node contains n/2 -1 values –Move the smallest of the remaining values, together with pointer, to the parent –2 nd node contains the remaining values

9 7 8 13 15
9
7 8
13 15

DBMS , A. Yaz c

19

B+ Tree: Insertion

l Example 1: Construct a B + tree for (1, 4, 7, 10, 17, 21, 31, 25, 19, 20, 28, 42) with n=4.

1 4 7
1
4
7
7 1 4 7 10
7
1
4
7
10
7 17 1 4 7 10 17 21
7
17
1
4
7
10
17
21

DBMS , A. Yaz c

20

B+ Tree: Insertion

B+ Tree Insertion

l 1, 4, 7, 10, 17, 21, 31, 25, 19, 20, 28, 42

7 17 25 1 4 7 10 17 21 25 31
7
17
25
1
4
7
10
17
21
25
31
17 7 20 25 1 4 7 10 17 19 20 21 25 31
17
7
20
25
1
4
7
10
17
19
20
21
25
31

DBMS , A. Yaz c

21

B+ Tree: Insertion

l 1, 4, 7, 10, 17, 21, 31, 25, 19, 20, 28, 42

17
17
Insertion l 1, 4, 7, 10, 17, 21, 31, 25, 19, 20, 28, 42 17 7
7
7
l 1, 4, 7, 10, 17, 21, 31, 25, 19, 20, 28, 42 17 7 20
20 25 31
20
25
31
1, 4, 7, 10, 17, 21, 31, 25, 19, 20, 28, 42 17 7 20 25
1, 4, 7, 10, 17, 21, 31, 25, 19, 20, 28, 42 17 7 20 25
1 4 7 10
1
4
7 10
1, 4, 7, 10, 17, 21, 31, 25, 19, 20, 28, 42 17 7 20 25
17 19
17
19
1, 4, 7, 10, 17, 21, 31, 25, 19, 20, 28, 42 17 7 20 25
31 42
31
42
20 21
20
21
25 28
25
28
1, 4, 7, 10, 17, 21, 31, 25, 19, 20, 28, 42 17 7 20 25

DBMS , A. Yaz c

22

B+ Tree: Insertion

l Example 2: n=3, insert 4 into the following B+Tree

9 10 7 8 Subtree Subtree C D 2 5 Leaf Leaf A B
9
10
7
8
Subtree
Subtree
C
D
2
5
Leaf
Leaf
A
B
9 7 10 4 8 C D 5 2 4 A B
9
7
10
4
8
C
D
5
2
4
A
B

DBMS , A. Yaz c

23

B+ Tree: Deletion

l Underflow

– When number of search-key values < n/2 -1

–Leaf Node

•Redistribute to sibling •Right node not less than left node •Replace the between-value in parent by their smallest value of the right node •Merge (contain too few entries) •Move all values, pointers to left node •Remove the between-value in parent

9 10 9 10 9 13
9
10
9
10
9
13

Delete 10

13 18
13
18
13 14 16
13
14 16
14 18
14 18
14 16
14 16

13

18

22

13 18 22
9 10 9 13 Delete 10 13 18 13 14 16 14 18 14 16 13
13 14
13
14
9 10
9
10

DBMS , A. Yaz c

18 22 9 13 14
18
22
9
13
14

24

B+ Tree: Deletion

–Non-Leaf Node

•Redistribute to sibling •Through parent •Right node not less than left node •Merge (contain too few entries) •Bring down parent •Move all values, pointers to left node •Delete the right node, and pointers in parent

9 10 9 10 9 13
9
10
9
10
9
13

Delete 10

13 18
13
18
14 15 16
14
15 16
14 18
14
18
15 16
15 16
13 18 22 9 10 14 16
13
18
22
9
10
14 16

DBMS , A. Yaz c

18 22 9 13 14 16
18
22
9
13
14 16

25

B+ Tree: Deletion

l Example 3: n=3, delete 3

5 20 3 Subtree A 1 3
5
20
3
Subtree
A
1
3

DBMS , A. Yaz c

20 Subtree 5
20
Subtree
5

A

B+ Tree: Deletion l Example 3: n=3, delete 3 5 20 3 Subtree A 1 3
1
1

26

B+ Tree: Deletion

l Example 4: Delete 28, 31, 21, 25, 19

20 7 17 25 31 50 1 4 7 10 17 19 20 21 25
20
7
17
25
31
50
1
4
7
10
17
19
20
21
25
28
31
42
20 7 17 25 50 1 4 7 10 17 19 20 21 25 31
20
7
17
25
50
1
4
7
10
17
19
20
21
25
31
42

DBMS , A. Yaz c

27

B+ Tree: Deletion

l Example 4: Delete 28, 31, 21, 25, 19

7 17 20 50
7 17
20
50

17

19

l Example 4: Delete 28, 31, 21, 25, 19 7 17 20 50 17 19 1
1 4 7 10
1
4
7 10
20 25 42
20
25
42
28, 31, 21, 25, 19 7 17 20 50 17 19 1 4 7 10 20
7 17 50 1 4 7 10 17 20 42
7
17
50
1
4
7
10
17
20
42

DBMS , A. Yaz c

28

B + -Tree File Organization

l

Index file degradation problem is solved by using B + - Tree indices. Data file degradation problem is solved by using B + -Tree File Organization.

l

The leaf nodes in a B + -tree file organization store records, instead of pointers.

l

Since records are larger than pointers, the maximum number of records that can be stored in a leaf node is less than the number of pointers in a nonleaf node.

l

Leaf nodes are still required to be half full.

l

Insertion and deletion are handled in the same way as insertion and deletion of entries in a B + -tree index.

DBMS , A. Yaz c

29

B + -Tree File Organization (Cont.)

B + -Tree File Organization (Cont.) # $ % & ' l Good space utilization important

# $ % & '

l

Good space utilization important since records use more space than pointers.

l

To improve space utilization, involve more sibling nodes in redistribution during splits and merges – Involving 2 siblings in redistribution (to avoid split / merge where possible) results in each node having at least entries

DBMS , A. Yaz c

30

Indexing Strings

l Variable length strings as keys

– Variable fanout

– Use space utilization as criterion for splitting, not number of pointers

l Prefix compression

– Key values at internal nodes can be prefixes of full key

l Keep enough characters to distinguish entries in the subtrees separated by the key value – E.g. “Silas” and “Silberschatz” can be separated by “Silb”

DBMS , A. Yaz c

31

B-Tree Index Files

n

( $ ) * + , $ +

n

( + , + $ *

n - '
n - '

l Nonleaf node – pointers B i are the bucket or file record pointers.

DBMS , A. Yaz c

32

B-Tree Index File Example

B-Tree Index File Example B-tree (above) and B+-tree (below) on same data DBMS , A. Yaz

B-tree (above) and B+-tree (below) on same data

B-Tree Index File Example B-tree (above) and B+-tree (below) on same data DBMS , A. Yaz

DBMS , A. Yaz c

33

B-Tree Index Files (Cont.)

l

Advantages of B-Tree indices:

– May use less tree nodes than a corresponding B + -Tree.

– Sometimes possible to find search-key value before reaching leaf node.

l

Disadvantages of B-Tree indices:

– Only small fraction of all search-key values are found early

– Non-leaf nodes are larger, so fan-out is reduced. Thus,

B-Trees typically have greater depth than corresponding

B

+ -Tree

– Insertion and deletion more complicated than in B + -Trees

– Implementation is harder than B + -Trees.

l

Typically, advantages of B-Trees do not outweigh disadvantages.

DBMS , A. Yaz c

34

Index Definition in SQL

l

Create an index create index <index-name> on <relation-name> (<attribute-list>) E.g.: create index b-index on branch(branch_name)

l

Use create unique index to indirectly specify and enforce the condition that the search key is a candidate key. – Not really required if SQL unique integrity constraint is supported

l

To drop an index drop index <index-name>

DBMS , A. Yaz c

35