Heim > Datenbank > MySQL-Tutorial > Hauptteil

dump浅析B树索引

WBOY
Freigeben: 2016-06-07 16:13:16
Original
1114 Leute haben es durchsucht

oracle中默认的索引类型是B树索引。还有位图索引,反向键索引,hash索引,基于函数的索引。 本篇主要介绍B树索引,通过转储分析。对于索引的扫描类型,索引的基本操作不做详细的介绍。 系统信息: [oracle@localhost ~]$ cat /etc/issue Enterprise Linux Ent

oracle中默认的索引类型是B树索引。还有位图索引,反向键索引,hash索引,基于函数的索引。
本篇主要介绍B树索引,通过转储分析。对于索引的扫描类型,索引的基本操作不做详细的介绍。
系统信息:
[oracle@localhost ~]$ cat /etc/issue
Enterprise Linux Enterprise Linux Server release 5.5 (Carthage)
Kernel \r on an \m

数据库版本:
SQL> select * from v$version where rownum =1 ;

BANNER
--------------------------------------------------------------------------------
Oracle Database 11g Enterprise Edition Release 11.2.0.1.0 - Production

SQL> show user;
USER 为 "HR"
SQL> desc tt;
名称 是否为空? 类型
----------------------------------------- -------- ----------------------------
ID NUMBER
NAME VARCHAR2(10)
SQL> select count(rownum) from tt;
COUNT(ROWNUM)
-------------
3670016
基于ID创建索引index_t

SQL> create index index_t on tt(id) tablespace users;

索引已创建。

SQL> select object_id from dba_objects where object_name=
2 'INDEX_T';
数据库中segment有数据段,索引段,undo段,它们和表名,索引名不是同一概念,但是名字是相同的。

OBJECT_ID
----------
76332
转储索引:
SQL> alter session set events 'immediate trace name treedump level 76332';
会话已更改。

----- begin tree dump
branch: 0x10038ab 16791723 (0: nrow: 15, level: 2)
branch: 0x100540b 16798731 (-1: nrow: 503, level: 1)
leaf: 0x10038ac 16791724 (-1: nrow: 512 rrow: 512)
leaf: 0x10038ad 16791725 (0: nrow: 512 rrow: 512)
leaf: 0x10038ae 16791726 (1: nrow: 512 rrow: 512)
leaf: 0x10038af 16791727 (2: nrow: 512 rrow: 512)
leaf: 0x10038b0 16791728 (3: nrow: 512 rrow: 512)
leaf: 0x10038b1 16791729 (4: nrow: 512 rrow: 512)
leaf: 0x10038b2 16791730 (5: nrow: 512 rrow: 512)
leaf: 0x10038b3 16791731 (6: nrow: 512 rrow: 512)
leaf: 0x10038b4 16791732 (7: nrow: 512 rrow: 512)
leaf: 0x10038b5 16791733 (8: nrow: 512 rrow: 512)
leaf: 0x10038b6 16791734 (9: nrow: 512 rrow: 512)
leaf: 0x10038b7 16791735 (10: nrow: 512 rrow: 512)
leaf: 0x10038b9 16791737 (11: nrow: 512 rrow: 512)
leaf: 0x10038ba 16791738 (12: nrow: 512 rrow: 512)
leaf: 0x10038bb 16791739 (13: nrow: 512 rrow: 512)
leaf: 0x10038bc 16791740 (14: nrow: 512 rrow: 512)
leaf: 0x10038bd 16791741 (15: nrow: 512 rrow: 512)
leaf: 0x10038be 16791742 (16: nrow: 512 rrow: 512)
leaf: 0x10038bf 16791743 (17: nrow: 512 rrow: 512)
leaf: 0x10038c0 16791744 (18: nrow: 512 rrow: 512)
leaf: 0x10038c1 16791745 (19: nrow: 512 rrow: 512)
leaf: 0x10038c2 16791746 (20: nrow: 512 rrow: 512)
leaf: 0x10038c3 16791747 (21: nrow: 512 rrow: 512)
leaf: 0x10038c4 16791748 (22: nrow: 512 rrow: 512)
leaf: 0x10038c5 16791749 (23: nrow: 512 rrow: 512)
leaf: 0x10038c6 16791750 (24: nrow: 512 rrow: 512)
leaf: 0x10038c7 16791751 (25: nrow: 512 rrow: 512)
leaf: 0x10038c9 16791753 (26: nrow: 512 rrow: 512)
leaf: 0x10038ca 16791754 (27: nrow: 512 rrow: 512)
leaf: 0x10038cb 16791755 (28: nrow: 512 rrow: 512)
leaf: 0x10038cc 16791756 (29: nrow: 512 rrow: 512)
leaf: 0x10038cd 16791757 (30: nrow: 512 rrow: 512)
leaf: 0x10038ce 16791758 (31: nrow: 512 rrow: 512)
leaf: 0x10038cf 16791759 (32: nrow: 512 rrow: 512)
leaf: 0x10038d0 16791760 (33: nrow: 512 rrow: 512)
leaf: 0x10038d1 16791761 (34: nrow: 512 rrow: 512)
leaf: 0x10038d2 16791762 (35: nrow: 512 rrow: 512)
leaf: 0x10038d3 16791763 (36: nrow: 512 rrow: 512)
leaf: 0x10038d4 16791764 (37: nrow: 512 rrow: 512)
leaf: 0x10038d5 16791765 (38: nrow: 512 rrow: 512)
leaf: 0x10038d6 16791766 (39: nrow: 512 rrow: 512)
leaf: 0x10038d7 16791767 (40: nrow: 512 rrow: 512)
leaf: 0x10038d9 16791769 (41: nrow: 512 rrow: 512)
leaf: 0x10038da 16791【本文来自鸿网互联 (http://www.68idc.cn)】770 (42: nrow: 512 rrow: 512)
leaf: 0x10038db 16791771 (43: nrow: 512 rrow: 512)
leaf: 0x10038dc 16791772 (44: nrow: 512 rrow: 512)
leaf: 0x10038dd 16791773 (45: nrow: 512 rrow: 512)
leaf: 0x10038de 16791774 (46: nrow: 512 rrow: 512)
...........................................................................
....
....
....
----- end tree dump
Nach dem Login kopieren

这是一棵平衡树,因为平衡树的查找效率很高,根节点到所有的叶子节点的高度相同。 branch表示的是根节点。以上选取了一部分,已经按从左向右拍好序了。
leaf: 0x10038ac 16791724 (-1: nrow: 512 rrow: 512) 我们选取这一列:

把十六进制,和十进制数相互转换:
SQL> select to_number('10038ac','xxxxxxxxxxxxxxxx') from dual;

TO_NUMBER('10038AC','XXXXXXXXXXXXXXXX')
---------------------------------------
16791724

SQL> select to_char('16791724','xxxxxxxxxxxxxxxx') from dual;


TO_CHAR('16791724','XXXXXXXXXXXXXX
----------------------------------
10038ac
Nach dem Login kopieren


我们利用oracle中提供的一个包可以求得索引所在的文件号,块号:
SQL> select dbms_utility.data_block_address_file(16791724) from dual;


DBMS_UTILITY.DATA_BLOCK_ADDRESS_FILE(16791724)
----------------------------------------------
4

SQL> select dbms_utility.data_block_address_block(16791724) from dual;


DBMS_UTILITY.DATA_BLOCK_ADDRESS_BLOCK(16791724)
-----------------------------------------------

14508
Nach dem Login kopieren



通过查视图dba_extends,索引存储在数据文件4,块14508在起始块范围内。
\
此时我们dump数据文件4,块14508:

SQL> alter system dump datafile 4 block 14508;


系统已更改。
row#0[8020] flag: ------, lock: 0, len=12
col 0; len 2; (2): c1 02
col 1; len 6; (6): 01 00 1b ab 00 01
row#1[8008] flag: ------, lock: 0, len=12
col 0; len 2; (2): c1 02
col 1; len 6; (6): 01 00 1b ab 00 03
row#2[7996] flag: ------, lock: 0, len=12
col 0; len 2; (2): c1 02
col 1; len 6; (6): 01 00 1b ab 00 08
row#3[7984] flag: ------, lock: 0, len=12
col 0; len 2; (2): c1 02
col 1; len 6; (6): 01 00 1b ab 00 0a
row#4[7972] flag: ------, lock: 0, len=12
col 0; len 2; (2): c1 02
col 1; len 6; (6): 01 00 1b ab 00 0f
row#5[7960] flag: ------, lock: 0, len=12
col 0; len 2; (2): c1 02
col 1; len 6; (6): 01 00 1b ab 00 11
row#6[7948] flag: ------, lock: 0, len=12
col 0; len 2; (2): c1 02
col 1; len 6; (6): 01 00 1b ab 00 16
row#7[7936] flag: ------, lock: 0, len=12
col 0; len 2; (2): c1 02
col 1; len 6; (6): 01 00 1b ab 00 18
row#8[7924] flag: ------, lock: 0, len=12
col 0; len 2; (2): c1 02
col 1; len 6; (6): 01 00 1b ab 00 1d
row#9[7912] flag: ------, lock: 0, len=12
col 0; len 2; (2): c1 02
col 1; len 6; (6): 01 00 1b ab 00 1f
row#10[7900] flag: ------, lock: 0, len=12
col 0; len 2; (2): c1 02
col 1; len 6; (6): 01 00 1b ab 00 24
row#11[7888] flag: ------, lock: 0, len=12
col 0; len 2; (2): c1 02
col 1; len 6; (6): 01 00 1b ab 00 26
row#12[7876] flag: ------, lock: 0, len=12
col 0; len 2; (2): c1 02
col 1; len 6; (6): 01 00 1b ab 00 2b
row#13[7864] flag: ------, lock: 0, len=12
col 0; len 2; (2): c1 02
col 1; len 6; (6): 01 00 1b ab 00 2d
row#14[7852] flag: ------, lock: 0, len=12
col 0; len 2; (2): c1 02
col 1; len 6; (6): 01 00 1b ab 00 32
row#15[7840] flag: ------, lock: 0, len=12
col 0; len 2; (2): c1 02
col 1; len 6; (6): 01 00 1b ab 00 34
row#16[7828] flag: ------, lock: 0, len=12
col 0; len 2; (2): c1 02
col 1; len 6; (6): 01 00 1b ab 00 39
row#17[7816] flag: ------, lock: 0, len=12
col 0; len 2; (2): c1 02
col 1; len 6; (6): 01 00 1b ab 00 3b
row#18[7804] flag: ------, lock: 0, len=12
col 0; len 2; (2): c1 02
col 1; len 6; (6): 01 00 1b ab 00 40
row#19[7792] flag: ------, lock: 0, len=12
col 0; len 2; (2): c1 02
col 1; len 6; (6): 01 00 1b ab 00 42
row#20[7780] flag: ------, lock: 0, len=12
...............................................................
Nach dem Login kopieren

.......
选取这一行为例:
row#0[8020] flag: ------, lock: 0, len=12
col 0; len 2; (2): c1 02
col 1; len 6; (6): 01 00 1b ab 00 01
col 0 表示第一列,长度为2,c1 02表示是多少呢?

SQL> select * from tt where rownum<4 order by id;


ID NAME
---------- --------------------
1 wO
2 wang
6 hong

SQL> select dump(1,16) from dual;


DUMP(1,16)
----------------------------------
Nach dem Login kopieren

Typ=2 Len=2: c1,2 ==>这里是c1 02,0省略了

这下清楚了吧,第一行第一列存储的是1,orale存储数据的方法很复杂。

col 1表示的是第二列,长度是6, 01 00 1b ab 00 01就是索引的值,是十六进制数,我们可以转化为二进制数:
00000001 00000000 00011011 10101011 00000000 00000001
-----------------------------------------------------------------------------------------------------------------------------------------------------------
00000001 00 ==>1x2x2=4 前10位表示了数据文件号
000000 00011011 10101011 ==>4096+2048+512+256+128+32+8+2+1=7083 这里的22位表示了块号
00000000 00000001==>1 16位代表着行号
此时排序并查rowid:
\
此时用第一列的rowid,
通过oracle提供的一个包,可以求出对象编号,文件号,块号:
执行如下图:
\
上面从索引存储的段以及数据段进行分析。
我们知道索引不一定会提高查询效率,往往乱建索引会严重影响查询效率,系统用不用索引,我们不能干预(但是dba可以手动改变),是oracle CBO选择的结果。
下面我们可以做一个小实验:
SQL> select count(rowid) from t;


COUNT(ROWID)
------------
4718644


SQL> select count(rowid) from t where id=1;


COUNT(ROWID)
------------
4718592


SQL> select count(rowid) from t where id=2;


COUNT(ROWID)
------------
26


SQL> select count(rowid) from t where id=3;


COUNT(ROWID)
------------
26


SQL> set autotrace traceonly;
SQL> select * from tt;
Nach dem Login kopieren



执行计划
----------------------------------------------------------
Plan hash value: 264906180


--------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 3556K| 67M| 1742 (2)| 00:00:21 |
| 1 | TABLE ACCESS FULL| TT | 3556K| 67M| 1742 (2)| 00:00:21 |
--------------------------------------------------------------------------


Note
-----
- dynamic sampling used for this statement (level=2)
Nach dem Login kopieren


此时是全表扫描读取,是多块读取,,这样读取比较快,如果此时用索引,则效率会低。
SQL> select * from tt where id=5;


执行计划
----------------------------------------------------------
Plan hash value: 3103123359


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


-------


| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Tim


e |


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


-------


| 0 | SELECT STATEMENT | | 5 | 100 | 4 (0)| 00:


00:01 |


| 1 | TABLE ACCESS BY INDEX ROWID| TT | 5 | 100 | 4 (0)| 00:


00:01 |


|* 2 | INDEX RANGE SCAN | INDEX_T | 5 | | 3 (0)| 00:


00:01 |


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


-------




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


2 - access("ID"=5)


Note
-----
- dynamic sampling used for this statement (level=2)
Nach dem Login kopieren



此时是索引读取。
SQL> select * from tt where id=1;


执行计划
----------------------------------------------------------
Plan hash value: 264906180


--------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)| Time |
--------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 2121K| 40M| 3644 (2)| 00:00:44 |
|* 1 | TABLE ACCESS FULL| TT | 2121K| 40M| 3644 (2)| 00:00:44 |
--------------------------------------------------------------------------


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


1 - filter("ID"=1)


Note
-----
- dynamic sampling used for this statement (level=2)
Nach dem Login kopieren


看到了吧,此时是全表扫描读取,数据库是很聪明的吧!


Verwandte Etiketten:
Quelle:php.cn
Erklärung dieser Website
Der Inhalt dieses Artikels wird freiwillig von Internetnutzern beigesteuert und das Urheberrecht liegt beim ursprünglichen Autor. Diese Website übernimmt keine entsprechende rechtliche Verantwortung. Wenn Sie Inhalte finden, bei denen der Verdacht eines Plagiats oder einer Rechtsverletzung besteht, wenden Sie sich bitte an admin@php.cn
Beliebte Tutorials
Mehr>
Neueste Downloads
Mehr>
Web-Effekte
Quellcode der Website
Website-Materialien
Frontend-Vorlage
Über uns Haftungsausschluss Sitemap
Chinesische PHP-Website:Online-PHP-Schulung für das Gemeinwohl,Helfen Sie PHP-Lernenden, sich schnell weiterzuentwickeln!