Question

我正在尝试查找未使用哪些索引以清除它们并减少SQL Server索引维护。

所以，我设法根据official documentation和其他文章（like this one）提取了两个查询：

第一个是基于sys.dm_db_index_usage_stats返回未使用的索引：

DECLARE @MinimumPageCount INT = 0

SELECT object_name(IDX.[object_id]) AS [Table]    
      ,IDX.[name] AS [Index] 
      ,PHS.[page_count] AS [Page_Count]
      ,CONVERT(DECIMAL(18,2), PHS.page_count * 8 / 1024.0) AS [Total Size (MB)]
      ,CONVERT(DECIMAL(18,2), PHS.avg_fragmentation_in_percent) AS [Frag %]
      ,PRS.row_count AS [Row Count]
      ,CONVERT(DECIMAL(18,2), (PHS.page_count * 8.0 * 1024) / PRS.row_count) AS [Index Size/Row (Bytes)]
      ,US.[user_updates]
      ,US.[last_user_update]
FROM [sys].[dm_db_index_usage_stats] US
INNER JOIN [sys].[indexes] IDX
    ON IDX.[index_id] = US.[index_id]
    AND IDX.[object_id] = US.[object_id]
INNER JOIN [sys].[dm_db_index_physical_stats] (DB_ID(),NULL,NULL,NULL,NULL) AS PHS
    ON PHS.[index_id] = US.[index_id] 
    AND PHS.[object_id] = US.[object_id]
INNER JOIN [sys].[dm_db_partition_stats] PRS
    ON PRS.[index_id] = US.[index_id]
    AND PRS.[object_id] = US.[object_id]  
WHERE 
    -- showing only not used indexes 
        US.[user_scans] = 0
    AND US.[user_seeks] = 0
    AND US.[user_lookups] = 0
    AND US.[system_scans] = 0
    AND US.[system_seeks] = 0
    AND US.[system_lookups] = 0
    -- ignore indexes with less than a certain number of pages of memory
    AND PHS.page_count > @MinimumPageCount
    -- exclude clustered and heaps tables
    AND US.[index_id] > 1
    --current db only 
    AND US.[database_id]= DB_ID()
ORDER BY [Page_Count] DESC

第二个是为sys.dm_db_index_usage_stats中没有出口的索引提取类似的信息，如下所示：

DECLARE @dbid INT =  DB_ID()

SELECT DB.Name AS [Database]
      ,OBJ.NAME AS [Table]
      ,IDX.NAME AS [Index]
      ,IDX.INDEX_ID
      ,PHS.page_count AS [Page Count]
      ,CONVERT(DECIMAL(18,2), PHS.page_count * 8 / 1024.0) AS [Total Index Size (MB)]
      ,CONVERT(DECIMAL(18,2), PHS.avg_fragmentation_in_percent) AS [Fragmentation (%)]
FROM [sys].[indexes] IDX
INNER JOIN SYS.OBJECTS OBJ 
    ON IDX.[object_id] = OBJ.[object_id]
LEFT JOIN [sys].[dm_db_index_physical_stats] (@dbid, NULL, NULL, NULL, NULL) PHS
    ON PHS.[object_id] = IDX.[object_id] 
    AND PHS.[index_id] = IDX.[index_id]
INNER JOIN [sys].[databases] DB
    ON DB.[database_id] = PHS.[database_id]
WHERE OBJ.[type] = 'U' -- Is User Table
    AND IDX.[is_primary_key] = 0
    AND IDX.[index_id] NOT IN 
    (
        SELECT US.INDEX_ID
        FROM [sys].[dm_db_index_usage_stats] US
        WHERE US.[object_id] = IDX.[object_id]
            AND IDX.[index_id] = US.[index_id]
            AND [database_id] = @dbid
    )
    AND IDX.[index_id] > 1
ORDER BY PHS.[page_count] DESC
        ,OBJ.[name]
        ,IDX.[index_id]
        ,IDX.[name] ASC

查看上面的查询，唯一的过滤是基于特定索引的sys.dm_db_index_usage_stats中是否有条目。

那么，是否存在一条规则，即如果没有条目，则不使用索引？

Answer 1

您提供的第一个查询搜索未使用的索引。 sys.dm_db_index_usage_stats函数具有[user_scans]，[user_seeks]，[user_lookups]，[system_scans]，[system_seeks]，[system_lookups]字段。如果它们都是0，则根据您自己的链接，这意味着索引从未使用过。

第二个查询给出了没有可用的使用统计信息的索引。根据您的链接，sys.dm_db_index_usage_stats函数的记录在每次重新启动时都设置为零，并在数据库分离或关闭时删除。因此，如果从数据库分离或关闭后从未使用过索引，那么就不会有使用统计行。

您已经提供了这个链接，但为了完整答案，我再次给它：http://technet.microsoft.com/en-us/library/ms188755.aspx

对于哪些索引，索引使用情况统计表中没有条目？

1 个答案: