索引質(zhì)量的高低對數(shù)據(jù)庫整體性能有著直接的影響。良好高質(zhì)量的索引使得數(shù)據(jù)庫性能得以數(shù)量級別的提升，而低效冗余的索引則使得數(shù)據(jù)庫性能緩慢如牛，即便是使用高檔的硬件配置。因此對于索引在設(shè)計之初需要經(jīng)過反復(fù)的測試與考量。那對于已經(jīng)置于生產(chǎn)環(huán)境中的數(shù)據(jù)庫，我們也可以通過查詢相關(guān)數(shù)據(jù)字典得到索引的質(zhì)量的高低，通過這個分析來指導(dǎo)如何改善索引的性能。下面給出了演示以及索引創(chuàng)建的基本指導(dǎo)原則，最后給出了索引質(zhì)量分析腳本。
1、查看索引質(zhì)量
|
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
|
--獲取指定schema或表上的索引質(zhì)量信息報告gx_adm@CABO3> @idx_qualityEnter value for input_owner: GX_ADMEnter value for input_tbname: CLIENT_TRADE_TBL -->如果我們省略具體的表名則會輸出整個schema的索引質(zhì)量報告 Table Table Index Data Blks Leaf Blks Clust IndexTable Rows Blocks Index Size MB per Key per Key Factor Quality------------------------- ------------ ---------- ------------------------- ------- --------- --------- ------------ -------------CLIENT_TRADE_TBL 6,318,035 278488 I_TDCL_ARC_STL_DATE_STOCK 62 312 13 171,017 5-Excellent I_TDCL_ARC_STL_DATE_CASH 62 318 13 174,599 5-Excellent I_TDCL_ARC_CANCEL_DATE 83 238 8 288,678 5-Excellent I_TDCL_ARC_INPUT_DATE 144 249 13 310,974 5-Excellent I_TDCL_ARC_TRADE_DATE 144 269 14 337,097 5-Excellent PK_CLIENT_TRADE_TBL 200 1 1 798,216 2-Good I_TDCL_ARC_GRP_REF_ID 144 1 1 811,468 2-Good UNI_TDCL_ARC_REF_ID 136 1 1 765,603 2-Good I_TDCL_ARC_CONTRACT_NUM 72 1 1 834,491 2-Good I_TDCL_ARC_SETTLED_DATE 61 299 5 380,699 1-Poor I_TDCL_ARC_ACC_NUM 184 624 3 3,899,446 1-Poor I_TDCL_ARC_PL_STK 176 218 1 4,348,804 1-Poor I_TDCL_ARC_INSTRU_ID 120 2,667 8 4,273,038 1-Poor--從上面的單表輸出的索引質(zhì)量可知，出現(xiàn)了4個處于Poor級別的索引，也就是說這些個索引具有較大的聚簇因子，幾乎接近于表上的行了--對于這幾個索引的質(zhì)量還應(yīng)結(jié)合該索引的使用頻率來考量該索引存在的必要性--對于聚簇因子，只能通過重新組織表上的數(shù)據(jù)來，以及調(diào)整相應(yīng)索引列的順序得以改善 --查詢單表上索引列的相關(guān)信息 gx_adm@CABO3> @idx_infoEnter value for owner: GX_ADMEnter value for table_name: CLIENT_TRADE_TBLTABLE_NAME INDEX_NAME CL_NAM CL_POS STATUS IDX_TYP DSCD------------------------- ------------------------------ -------------------- ------ -------- --------------- ----CLIENT_TRADE_TBL I_TDCL_ARC_ACC_NUM ACC_NUM 1 VALID NORMAL ASC I_TDCL_ARC_CANCEL_DATE CANCEL_DATE 1 VALID NORMAL ASC I_TDCL_ARC_CONTRACT_NUM CONTRACT_NUM 1 VALID NORMAL ASC I_TDCL_ARC_GRP_REF_ID GRP_REF_ID 1 VALID NORMAL ASC I_TDCL_ARC_INPUT_DATE INPUT_DATE 1 VALID NORMAL ASC I_TDCL_ARC_INSTRU_ID INSTRU_ID 1 VALID NORMAL ASC I_TDCL_ARC_PL_STK STOCK_CD 1 VALID NORMAL ASC I_TDCL_ARC_PL_STK PL_CD 2 VALID NORMAL ASC I_TDCL_ARC_SETTLED_DATE SETTLED_DATE 1 VALID NORMAL ASC I_TDCL_ARC_STL_DATE_CASH STL_DATE_CASH 1 VALID NORMAL ASC I_TDCL_ARC_STL_DATE_STOCK STL_DATE_STOCK 1 VALID NORMAL ASC I_TDCL_ARC_TRADE_DATE TRADE_DATE 1 VALID NORMAL ASC PK_CLIENT_TRADE_TBL BUSINESS_DATE 1 VALID NORMAL ASC PK_CLIENT_TRADE_TBL REF_ID 2 VALID NORMAL ASC UNI_TDCL_ARC_REF_ID REF_ID 1 VALID NORMAL ASC --從上面的查詢結(jié)果可知，當(dāng)前表TRADE_CLIENT_TBL上含有13個索引，應(yīng)該來說該表索引存在一定冗余。--大多數(shù)情況下，單表上6-7個索引是比較理想的。過多的索引導(dǎo)致過大的資源開銷，以及降低DML性能。 |
2、索引創(chuàng)建的基本指導(dǎo)原則
索引的創(chuàng)建應(yīng)遵循精而少的原則
收集表上所有查詢的各種不同組合，找出具有最佳離散度的列(或主鍵列等)創(chuàng)建單索引
對于頻繁讀取而缺乏比較理想離散值的列為其創(chuàng)建組合索引
對于組合索引應(yīng)考慮下列因素來制定合理的索引列順序，以下優(yōu)先級別由高到低來作為索引的前導(dǎo)列，第二列等等
列被使用的頻率
該列是否經(jīng)常使用“ = ”作為常用查詢條件
列上的離散度
組合列經(jīng)常按何種順序排序
哪些列會作為附件性列被添加
3、索引質(zhì)量分析腳本
|
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
|
--script name: idx_quality.sql --Author : Leshami --Blog: http://blog.csdn.net/leshami --index quality retrievalSET LINESIZE 145SET PAGESIZE 1000SET VERIFY OFFCLEAR COMPUTESCLEAR BREAKSBREAK ON table_name ON num_rows ON blocksCOLUMN owner FORMAT a14 HEADING 'Index owner'COLUMN table_name FORMAT a25 HEADING 'Table'COLUMN index_name FORMAT a25 HEADING 'Index'COLUMN num_rows FORMAT 999G999G990 HEADING 'Table|Rows'COLUMN MB FORMAT 9G990 HEADING 'Index|Size MB'COLUMN blocks HEADING 'Table|Blocks'COLUMN num_blocks FORMAT 9G990 HEADING 'Data|Blocks'COLUMN avg_data_blocks_per_key FORMAT 999G990 HEADING 'Data Blks|per Key'COLUMN avg_leaf_blocks_per_key FORMAT 999G990 HEADING 'Leaf Blks|per Key'COLUMN clustering_factor FORMAT 999G999G990 HEADING 'Clust|Factor'COLUMN Index_Quality FORMAT A13 HEADING 'Index|Quality'--SPOOL index_quality SELECT i.table_name, t.num_rows, t.blocks, i.index_name, o.bytes / 1048576 mb, i.avg_data_blocks_per_key, i.avg_leaf_blocks_per_key, i.clustering_factor, CASE WHEN NVL (i.clustering_factor, 0) = 0 THEN '0-No Stats' WHEN NVL (t.num_rows, 0) = 0 THEN '0-No Stats' WHEN (ROUND (i.clustering_factor / t.num_rows * 100)) < 6 THEN '5-Excellent' WHEN (ROUND (i.clustering_factor / t.num_rows * 100)) BETWEEN 7 AND 11 THEN '4-Very Good' WHEN (ROUND (i.clustering_factor / t.num_rows * 100)) BETWEEN 12 AND 15 THEN '2-Good' WHEN (ROUND (i.clustering_factor / t.num_rows * 100)) BETWEEN 16 AND 25 THEN '2-Fair' ELSE '1-Poor' END index_quality FROM dba_indexes i, dba_segments o, dba_tables t WHERE -- i.index_name LIKE UPPER ('%&&1%') AND i.owner = t.owner AND i.table_name = t.table_name AND i.owner = o.owner AND i.index_name = o.segment_name AND t.owner = UPPER('&input_owner') AND t.table_name LIKE UPPER('%&input_tbname%')ORDER BY table_name, num_rows, blocks, index_quality DESC;--SPOOL OFF;===========================================================================================--script name: idx_info.sql --get the index column information by specified tableset linesize 180col cl_nam format a20col table_name format a25col cl_pos format 9col idx_typ format a15SELECT b.table_name, a.index_name, a.column_name cl_nam, a.column_position cl_pos, b.status, b.index_type idx_typ, a.descend dscdFROM dba_ind_columns a, dba_indexes bWHERE a.index_name = b.index_name AND owner = upper('&owner') AND a.table_name LIKE upper('%&table_name%')ORDER BY 2, 4; |
