虚拟机测试等值索引扫描成本

SQL> create index test_idx1 on test(owner);

BEGIN
DBMS_STATS.GATHER_TABLE_STATS(ownname => 'TEST',
tabname => 'TEST',
estimate_percent => 100,
method_opt => 'for all columns size 1',
degree => DBMS_STATS.AUTO_DEGREE,
cascade=>TRUE
);
END;

SQL> select owner,blocks from dba_tables where owner='TEST' and table_name='TEST';

OWNER                              BLOCKS
------------------------------ ----------
TEST                                32910

SQL>  select count(distinct dbms_rowid.rowid_block_number(rowid)) from test;

COUNT(DISTINCTDBMS_ROWID.ROWID_BLOCK_NUMBER(ROWID))
---------------------------------------------------
                                              23563

SQL> show parameter multi

NAME                                 TYPE        VALUE
------------------------------------ ----------- ------------------------------
db_file_multiblock_read_count        integer     16
parallel_adaptive_multi_user         boolean     TRUE


SQL> select /*+ index(test test_idx1)*/ * from test where owner='SYS';

已选择986880行。


执行计划
----------------------------------------------------------
Plan hash value: 2624864549

-----------------------------------------------------------------------------------------
| Id  | Operation                   | Name      | Rows  | Bytes | Cost (%CPU)| Time     |
-----------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT            |           | 77444 |  7336K|  2241   (1)| 00:00:27 |
|   1 |  TABLE ACCESS BY INDEX ROWID| TEST      | 77444 |  7336K|  2241   (1)| 00:00:27 |
|*  2 |   INDEX RANGE SCAN          | TEST_IDX1 | 77444 |       |   185   (1)| 00:00:03 |
-----------------------------------------------------------------------------------------

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

   2 - access("OWNER"='SYS')


统计信息
----------------------------------------------------------
          1  recursive calls
          0  db block gets
     169962  consistent gets
      27475  physical reads
     915836  redo size
  108497026  bytes sent via SQL*Net to client
     724117  bytes received via SQL*Net from client
      65793  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
     986880  rows processed
成本为2241


SQL> select leaf_blocks,blevel,clustering_factor from dba_indexes where owner='TEST' and  index_name='TEST_IDX1';

LEAF_BLOCKS     BLEVEL CLUSTERING_FACTOR
----------- ---------- -----------------
       5452          2             61635


LEAF_BLOCKS 叶子块 5452个

BLEVEL  索引高度-1


SQL> select leaf_blocks,blevel,clustering_factor from dba_indexes where owner='TEST' and  index_name='TEST_IDX1';

LEAF_BLOCKS     BLEVEL CLUSTERING_FACTOR
----------- ---------- -----------------
       5452          2             61635

SQL> set linesize 200
SQL> select b.num_rows,
  2  a.density * (b.num_rows - a.num_nulls) ,
  3             a.num_distinct,
  4             a.num_nulls,
  5             utl_raw.cast_to_number(high_value) high_value,
  6             utl_raw.cast_to_number(low_value) low_value,
  7             (b.num_rows - a.num_nulls) "NUM_ROWS-NUM_NULLS",
  8             utl_raw.cast_to_number(high_value) -
  9             utl_raw.cast_to_number(low_value) "HIGH_VALUE-LOW_VALUE"
 10        from dba_tab_col_statistics a, dba_tables b
 11      where a.owner = b.owner
 12        and a.table_name = b.table_name
 13        and a.owner = 'TEST'
 14        and a.table_name = upper('TEST')
 15        and a.column_name = 'OWNER'; 

  NUM_ROWS A.DENSITY*(B.NUM_ROWS-A.NUM_NULLS) NUM_DISTINCT  NUM_NULLS HIGH_VALUE  LOW_VALUE NUM_ROWS-NUM_NULLS HIGH_VALUE-LOW_VALUE
---------- ---------------------------------- ------------ ---------- ---------- ---------- ------------------ --------------------
   2323328                                 .5           30          0                                  2323328           .000021321




单个谓词过滤：
=  基数计算公式  选择性计算为：1/num_distinct*(num_rows-num_nulls),如果有直方图，基数计算公式=(num_rows-num_nulls)*density



SQL> select a.column_name,
  2             b.num_rows,
  3             a.num_distinct Cardinality,
  4              round(a.num_distinct / b.num_rows * 100, 2) selectivity,
  5             a.histogram,
  6            a.num_buckets
  7        from dba_tab_col_statistics a, dba_tables b
  8       where a.owner = b.owner
  9         and a.table_name = b.table_name
 10        and a.owner = 'TEST'
 11       and a.table_name = 'TEST'
 12       and a.column_name='OWNER';

COLUMN_NAME                      NUM_ROWS CARDINALITY SELECTIVITY HISTOGRAM       NUM_BUCKETS
------------------------------ ---------- ----------- ----------- --------------- -----------
OWNER                             2323328          30           0 FREQUENCY                30

索引扫描的计算公式如下：
cost =  
 blevel +  
 celiling(leaf_blocks *effective index selectivity) +  
 celiling(clustering_factor * effective table selectivity)


选择性为:1/30

SQL> select 2+ceil(5452 *(1/30))+ceil(61635*(1/30)) from dual;

2+CEIL(5452*(1/30))+CEIL(61635*(1/30))
--------------------------------------
                                  2239

接近成本2241