Question

SQL Server 2016。

Table1 is (Col11, Col12, Col13)
Table2 is (Col21, Col22, Col23)

Col11和Col21分别是PK，两个索引都是聚类的 Col22引用Col11并编入索引 Col12也被编入索引。

查询：

SELECT count(*)  
FROM Table1 t1
JOIN Table2 t2 ON t2.col22 = t1.col11
WHERE t1.col12 = <value>

执行计划使用col12上的索引，但不使用col22上的索引（它使用与Table2不同的索引，在一个甚至不是查询的一部分的列上）。对于使用散列或嵌套循环连接的两个计划都是如此。

这种行为的潜在原因是什么以及如何解决？两个表都是〜1亿行，t1.col12的选择率约为2％。

对于那些想要剧本的人：

CREATE TABLE dbo.Table1 (
  Col11 BIGINT IDENTITY (1,1),
  Col12 BIGINT, 
  Col13 VARCHAR(100)
)

CREATE TABLE dbo.Table2 (
  Col21 BIGINT IDENTITY (1,1),
  Col22 BIGINT, 
  Col23 INTEGER
)

ALTER TABLE dbo.Table1 ADD 
  CONSTRAINT PK_Table1 PRIMARY KEY (
    Col11 ASC
  ) WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, 
          SORT_IN_TEMPDB = OFF, IGNORE_DUP_KEY = OFF, ONLINE = OFF,
          ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON)

ALTER TABLE dbo.Table2 ADD
  CONSTRAINT PK_Table2 PRIMARY KEY (
    Col21 ASC
  ) WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, 
          SORT_IN_TEMPDB = OFF, IGNORE_DUP_KEY = OFF, ONLINE = OFF, 
          ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON)

ALTER TABLE dbo.Table2 WITH NOCHECK ADD
  CONSTRAINT FK_Table2_Table1 FOREIGN KEY(Col22)
    REFERENCES dbo.Table1 (Col11)
    ON DELETE CASCADE

CREATE NONCLUSTERED INDEX IX_Table1_1 ON dbo.Table1 (Col12 ASC)
  WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, 
        SORT_IN_TEMPDB = OFF, DROP_EXISTING = OFF, ONLINE = OFF,
        ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON)

CREATE NONCLUSTERED INDEX IX_Table2_1 ON dbo.Table2 (Col22 ASC)
  WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, 
        SORT_IN_TEMPDB = OFF, DROP_EXISTING = OFF, ONLINE = OFF,
        ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON)

CREATE NONCLUSTERED INDEX IX_Table2_2 ON dbo.Table2 (Col23 ASC)
  WITH (PAD_INDEX = OFF, STATISTICS_NORECOMPUTE = OFF, 
        SORT_IN_TEMPDB = OFF, DROP_EXISTING = OFF, ONLINE = OFF,
        ALLOW_ROW_LOCKS = ON, ALLOW_PAGE_LOCKS = ON)

为什么优化器不使用索引（sql server）

0 个答案: