是否有人知道可以检测整个数据库中的冗余索引的T-SQL脚本？表中冗余索引的示例如下：
Index 1: 'ColumnA', 'ColumnB', 'ColumnC'
Index 2: 'ColumnA', 'ColumnB'
忽略其他注意事项，例如列宽和覆盖索引，索引2将是多余的。
感谢。
答案 0 :(得分：11)
有些情况下冗余不成立。例如，假设ColumnC是一个huuge字段，但有时您必须快速检索它。您的index 1不需要键查找：
select ColumnC from YourTable where ColumnnA = 12
另一方面，index 2要小得多，因此可以在内存中读取需要索引扫描的查询：
select * from YourTable where ColumnnA like '%hello%'
所以他们并不是多余的。
如果您不相信我的上述论点，您可以找到“冗余”索引，例如：
;with ind as (
select a.object_id
, a.index_id
, cast(col_list.list as varchar(max)) as list
from (
select distinct object_id
, index_id
from sys.index_columns
) a
cross apply
(
select cast(column_id as varchar(16)) + ',' as [text()]
from sys.index_columns b
where a.object_id = b.object_id
and a.index_id = b.index_id
for xml path(''), type
) col_list (list)
)
select object_name(a.object_id) as TableName
, asi.name as FatherIndex
, bsi.name as RedundantIndex
from ind a
join sys.sysindexes asi
on asi.id = a.object_id
and asi.indid = a.index_id
join ind b
on a.object_id = b.object_id
and a.object_id = b.object_id
and len(a.list) > len(b.list)
and left(a.list, LEN(b.list)) = b.list
join sys.sysindexes bsi
on bsi.id = b.object_id
and bsi.indid = b.index_id
为用户带来蛋糕，以防“性能”意外降低： - ）
答案 1 :(得分：1)
受Paul Nielsen的启发，我写了这个查询来查找/区分：
并记录他们的用法
（人们可能也想使用is_descending_key，但我不需要它。）
WITH IndexColumns AS
(
SELECT I.object_id AS TableObjectId, OBJECT_SCHEMA_NAME(I.object_id) + '.' + OBJECT_NAME(I.object_id) AS TableName, I.index_id AS IndexId, I.name AS IndexName
, (IndexUsage.user_seeks + IndexUsage.user_scans + IndexUsage.user_lookups) AS IndexUsage
, IndexUsage.user_updates AS IndexUpdates
, (SELECT CASE is_included_column WHEN 1 THEN NULL ELSE column_id END AS [data()]
FROM sys.index_columns AS IndexColumns
WHERE IndexColumns.object_id = I.object_id
AND IndexColumns.index_id = I.index_id
ORDER BY index_column_id, column_id
FOR XML PATH('')
) AS ConcIndexColumnNrs
,(SELECT CASE is_included_column WHEN 1 THEN NULL ELSE COL_NAME(I.object_id, column_id) END AS [data()]
FROM sys.index_columns AS IndexColumns
WHERE IndexColumns.object_id = I.object_id
AND IndexColumns.index_id = I.index_id
ORDER BY index_column_id, column_id
FOR XML PATH('')
) AS ConcIndexColumnNames
,(SELECT CASE is_included_column WHEN 1 THEN column_id ELSE NULL END AS [data()]
FROM sys.index_columns AS IndexColumns
WHERE IndexColumns.object_id = I.object_id
AND IndexColumns.index_id = I.index_id
ORDER BY column_id
FOR XML PATH('')
) AS ConcIncludeColumnNrs
,(SELECT CASE is_included_column WHEN 1 THEN COL_NAME(I.object_id, column_id) ELSE NULL END AS [data()]
FROM sys.index_columns AS IndexColumns
WHERE IndexColumns.object_id = I.object_id
AND IndexColumns.index_id = I.index_id
ORDER BY column_id
FOR XML PATH('')
) AS ConcIncludeColumnNames
FROM sys.indexes AS I
LEFT OUTER JOIN sys.dm_db_index_usage_stats AS IndexUsage
ON IndexUsage.object_id = I.object_id
AND IndexUsage.index_id = I.index_id
AND IndexUsage.Database_id = db_id()
)
SELECT
C1.TableName
, C1.IndexName AS 'Index1'
, C2.IndexName AS 'Index2'
, CASE WHEN (C1.ConcIndexColumnNrs = C2.ConcIndexColumnNrs) AND (C1.ConcIncludeColumnNrs = C2.ConcIncludeColumnNrs) THEN 'Exact duplicate'
WHEN (C1.ConcIndexColumnNrs = C2.ConcIndexColumnNrs) THEN 'Different includes'
ELSE 'Overlapping columns' END
-- , C1.ConcIndexColumnNrs
-- , C2.ConcIndexColumnNrs
, C1.ConcIndexColumnNames
, C2.ConcIndexColumnNames
-- , C1.ConcIncludeColumnNrs
-- , C2.ConcIncludeColumnNrs
, C1.ConcIncludeColumnNames
, C2.ConcIncludeColumnNames
, C1.IndexUsage
, C2.IndexUsage
, C1.IndexUpdates
, C2.IndexUpdates
, 'DROP INDEX ' + C2.IndexName + ' ON ' + C2.TableName AS Drop2
, 'DROP INDEX ' + C1.IndexName + ' ON ' + C1.TableName AS Drop1
FROM IndexColumns AS C1
INNER JOIN IndexColumns AS C2
ON (C1.TableObjectId = C2.TableObjectId)
AND (
-- exact: show lower IndexId as 1
(C1.IndexId < C2.IndexId
AND C1.ConcIndexColumnNrs = C2.ConcIndexColumnNrs
AND C1.ConcIncludeColumnNrs = C2.ConcIncludeColumnNrs)
-- different includes: show longer include as 1
OR (C1.ConcIndexColumnNrs = C2.ConcIndexColumnNrs
AND LEN(C1.ConcIncludeColumnNrs) > LEN(C2.ConcIncludeColumnNrs))
-- overlapping: show longer index as 1
OR (C1.IndexId <> C2.IndexId
AND C1.ConcIndexColumnNrs <> C2.ConcIndexColumnNrs
AND C1.ConcIndexColumnNrs like C2.ConcIndexColumnNrs + ' %')
)
ORDER BY C1.TableName, C1.ConcIndexColumnNrs
答案 2 :(得分：1)
我创建了以下查询，它为我提供了很多识别重复和近似重复索引的好信息。它还包括其他信息，例如索引占用的内存页数，这使我可以为更大的索引提供更高的优先级。它显示了索引的列以及包含的列，因此我可以看到是否有两个索引几乎完全相同，并且所包含的列中只有轻微的变化。
WITH IndexSummary AS
(
SELECT DISTINCT sys.objects.name AS [Table Name],
sys.indexes.name AS [Index Name],
SUBSTRING((SELECT ', ' + sys.columns.Name as [text()]
FROM sys.columns
INNER JOIN sys.index_columns
ON sys.index_columns.column_id = sys.columns.column_id
AND sys.index_columns.object_id = sys.columns.object_id
WHERE sys.index_columns.index_id = sys.indexes.index_id
AND sys.index_columns.object_id = sys.indexes.object_id
AND sys.index_columns.is_included_column = 0
ORDER BY sys.columns.name
FOR XML Path('')), 2, 10000) AS [Indexed Column Names],
ISNULL(SUBSTRING((SELECT ', ' + sys.columns.Name as [text()]
FROM sys.columns
INNER JOIN sys.index_columns
ON sys.index_columns.column_id = sys.columns.column_id
AND sys.index_columns.object_id = sys.columns.object_id
WHERE sys.index_columns.index_id = sys.indexes.index_id
AND sys.index_columns.object_id = sys.indexes.object_id
AND sys.index_columns.is_included_column = 1
ORDER BY sys.columns.name
FOR XML Path('')), 2, 10000), '') AS [Included Column Names],
sys.indexes.index_id, sys.indexes.object_id
FROM sys.indexes
INNER JOIN SYS.index_columns
ON sys.indexes.index_id = SYS.index_columns.index_id
AND sys.indexes.object_id = sys.index_columns.object_id
INNER JOIN sys.objects
ON sys.OBJECTS.object_id = SYS.indexES.object_id
WHERE sys.objects.type = 'U'
)
SELECT IndexSummary.[Table Name],
IndexSummary.[Index Name],
IndexSummary.[Indexed Column Names],
IndexSummary.[Included Column Names],
PhysicalStats.page_count as [Page Count],
CONVERT(decimal(18,2), PhysicalStats.page_count * 8 / 1024.0) AS [Size (MB)],
CONVERT(decimal(18,2), PhysicalStats.avg_fragmentation_in_percent) AS [Fragment %]
FROM IndexSummary
INNER JOIN sys.dm_db_index_physical_stats (DB_ID(), NULL, NULL, NULL, NULL)
AS PhysicalStats
ON PhysicalStats.index_id = IndexSummary.index_id
AND PhysicalStats.object_id = IndexSummary.object_id
WHERE (SELECT COUNT(*) as Computed
FROM IndexSummary Summary2
WHERE Summary2.[Table Name] = IndexSummary.[Table Name]
AND Summary2.[Indexed Column Names] = IndexSummary.[Indexed Column Names]) > 1
ORDER BY [Table Name], [Index Name], [Indexed Column Names], [Included Column Names]
查询结果如下：
Table Name Index Indexed Cols Included Cols Pages Size (MB) Frag %
My_Table Indx_1 Col1 Col2, Col3 123 0.96 8.94
My_Table Indx_2 Col1 Col2, Col3 123 0.96 8.94
有关完整说明，请参阅 Identifying Duplicate or Redundant Indexes in SQL Server 。
答案 3 :(得分：1)
尝试下面的脚本来显示未使用的索引，希望它有所帮助
/****************************************************************
Description: Script to show Unused Indexes using DMVs
****************************************************************/
SELECT TOP 100
o.name AS ObjectName
, i.name AS IndexName
, i.index_id AS IndexID
, dm_ius.user_seeks AS UserSeek
, dm_ius.user_scans AS UserScans
, dm_ius.user_lookups AS UserLookups
, dm_ius.user_updates AS UserUpdates
, p.TableRows
, 'DROP INDEX ' + QUOTENAME(i.name)
+ ' ON ' + QUOTENAME(s.name) + '.' + QUOTENAME(OBJECT_NAME(dm_ius.object_id)) as 'drop statement'
FROM sys.dm_db_index_usage_stats dm_ius
INNER JOIN sys.indexes i ON i.index_id = dm_ius.index_id AND dm_ius.object_id = i.object_id
INNER JOIN sys.objects o on dm_ius.object_id = o.object_id
INNER JOIN sys.schemas s on o.schema_id = s.schema_id
INNER JOIN (SELECT SUM(p.rows) TableRows, p.index_id, p.object_id
FROM sys.partitions p GROUP BY p.index_id, p.object_id) p
ON p.index_id = dm_ius.index_id AND dm_ius.object_id = p.object_id
WHERE OBJECTPROPERTY(dm_ius.object_id,'IsUserTable') = 1
AND dm_ius.database_id = DB_ID()
AND i.type_desc = 'nonclustered'
AND i.is_primary_key = 0
AND i.is_unique_constraint = 0
ORDER BY (dm_ius.user_seeks + dm_ius.user_scans + dm_ius.user_lookups) ASC
GO
答案 4 :(得分：0)
我刚刚阅读了一些MSDN博客，注意到script to do this并记住了这个问题。
我并没有费心地与Andomar's一起测试，看看是否有人比另一方有任何特别的好处。
我可能会对两者做出一项修正，但在评估冗余时会考虑两个索引的大小。
修改
上的金伯利·特里普的帖子