有关查询和执行计划的DMV

转自:http://www.cnblogs.com/CareySon/archive/2012/05/17/2506035.html

查看被缓存的查询计划

SET TRANSACTION ISOLATION LEVEL READ UNCOMMITTED 
SELECT TOP 20 
    st.text AS [SQL] 
    , cp.cacheobjtype 
    , cp.objtype 
    , COALESCE(DB_NAME(st.dbid), 
        DB_NAME(CAST(pa.value AS INT))+'*', 
        'Resource') AS [DatabaseName] 
    , cp.usecounts AS [Plan usage] 
    , qp.query_plan 
FROM sys.dm_exec_cached_plans cp                       
CROSS APPLY sys.dm_exec_sql_text(cp.plan_handle) st 
CROSS APPLY sys.dm_exec_query_plan(cp.plan_handle) qp 
OUTER APPLY sys.dm_exec_plan_attributes(cp.plan_handle) pa 
WHERE pa.attribute = 'dbid' 
  AND st.text LIKE '%这里是查询语句包含的内容%'  

 

结果是:

1

 

可以根据查询字段来根据关键字查看缓冲的查询计划。

 

查看某一查询是如何使用查询计划的

 

SET TRANSACTION ISOLATION LEVEL READ UNCOMMITTED 
SELECT TOP 20 
  SUBSTRING (qt.text,(qs.statement_start_offset/2) + 1,       
  ((CASE WHEN qs.statement_end_offset = -1 
    THEN LEN(CONVERT(NVARCHAR(MAX), qt.text)) * 2 
      ELSE qs.statement_end_offset 
    END - qs.statement_start_offset)/2) + 1) AS [Individual Query] 
  , qt.text AS [Parent Query] 
  , DB_NAME(qt.dbid) AS DatabaseName 
  , qp.query_plan 
FROM sys.dm_exec_query_stats qs 
CROSS APPLY sys.dm_exec_sql_text(qs.sql_handle) as qt 
CROSS APPLY sys.dm_exec_query_plan(qs.plan_handle) qp 
WHERE SUBSTRING (qt.text,(qs.statement_start_offset/2) + 1, 
  ((CASE WHEN qs.statement_end_offset = -1 
    THEN LEN(CONVERT(NVARCHAR(MAX), qt.text)) * 2 
    ELSE qs.statement_end_offset 
    END - qs.statement_start_offset)/2) + 1) 
LIKE '%指定查询包含的字段%'  

 

结果是:

2

 

查看数据库中跑的最慢的前20个查询以及它们的执行计划

 

SET TRANSACTION ISOLATION LEVEL READ UNCOMMITTED 
SELECT TOP 20 
  CAST(qs.total_elapsed_time / 1000000.0 AS DECIMAL(28, 2)) 
                                     AS [Total Duration (s)] 
  , CAST(qs.total_worker_time * 100.0 / qs.total_elapsed_time 
                               AS DECIMAL(28, 2)) AS [% CPU] 
  , CAST((qs.total_elapsed_time - qs.total_worker_time)* 100.0 / 
        qs.total_elapsed_time AS DECIMAL(28, 2)) AS [% Waiting] 
  , qs.execution_count 
  , CAST(qs.total_elapsed_time / 1000000.0 / qs.execution_count 
                AS DECIMAL(28, 2)) AS [Average Duration (s)] 
  , SUBSTRING (qt.text,(qs.statement_start_offset/2) + 1,      
    ((CASE WHEN qs.statement_end_offset = -1 
      THEN LEN(CONVERT(NVARCHAR(MAX), qt.text)) * 2 
      ELSE qs.statement_end_offset 
      END - qs.statement_start_offset)/2) + 1) AS [Individual Query 
  , qt.text AS [Parent Query] 
  , DB_NAME(qt.dbid) AS DatabaseName 
  , qp.query_plan 
FROM sys.dm_exec_query_stats qs 
CROSS APPLY sys.dm_exec_sql_text(qs.sql_handle) as qt 
CROSS APPLY sys.dm_exec_query_plan(qs.plan_handle) qp 
WHERE qs.total_elapsed_time > 0 
ORDER BY qs.total_elapsed_time DESC                     

3

 

查看数据库中哪个查询最耗费资源有助于你解决问题

 

被阻塞时间最长的前20个查询以及它们的执行计划

 

SET TRANSACTION ISOLATION LEVEL READ UNCOMMITTED 
SELECT TOP 20 
  CAST((qs.total_elapsed_time - qs.total_worker_time) /      
        1000000.0 AS DECIMAL(28,2)) AS [Total time blocked (s)] 
  , CAST(qs.total_worker_time * 100.0 / qs.total_elapsed_time 
        AS DECIMAL(28,2)) AS [% CPU] 
  , CAST((qs.total_elapsed_time - qs.total_worker_time)* 100.0 / 
        qs.total_elapsed_time AS DECIMAL(28, 2)) AS [% Waiting] 
  , qs.execution_count 
  , CAST((qs.total_elapsed_time  - qs.total_worker_time) / 1000000.0 
    / qs.execution_count AS DECIMAL(28, 2)) AS [Blocking average (s)] 
  , SUBSTRING (qt.text,(qs.statement_start_offset/2) + 1,     
  ((CASE WHEN qs.statement_end_offset = -1 
    THEN LEN(CONVERT(NVARCHAR(MAX), qt.text)) * 2 
    ELSE qs.statement_end_offset 
    END - qs.statement_start_offset)/2) + 1) AS [Individual Query] 
  , qt.text AS [Parent Query] 
  , DB_NAME(qt.dbid) AS DatabaseName 
  , qp.query_plan 
FROM sys.dm_exec_query_stats qs 
CROSS APPLY sys.dm_exec_sql_text(qs.sql_handle) as qt 
CROSS APPLY sys.dm_exec_query_plan(qs.plan_handle) qp 
WHERE qs.total_elapsed_time > 0 
ORDER BY [Total time blocked (s)] DESC                       

结果如图:

4

 

找出这类查询也是数据库调优的必须品

 

最耗费CPU的前20个查询以及它们的执行计划

 

SET TRANSACTION ISOLATION LEVEL READ UNCOMMITTED 
SELECT TOP 20 
  CAST((qs.total_worker_time) / 1000000.0 AS DECIMAL(28,2)) 
                                           AS [Total CPU time (s)] 
  , CAST(qs.total_worker_time * 100.0 / qs.total_elapsed_time 
                                      AS DECIMAL(28,2)) AS [% CPU] 
  , CAST((qs.total_elapsed_time - qs.total_worker_time)* 100.0 / 
           qs.total_elapsed_time AS DECIMAL(28, 2)) AS [% Waiting] 
             , qs.execution_count 
  , CAST((qs.total_worker_time) / 1000000.0 
    / qs.execution_count AS DECIMAL(28, 2)) AS [CPU time average (s)] 
  , SUBSTRING (qt.text,(qs.statement_start_offset/2) + 1,      
    ((CASE WHEN qs.statement_end_offset = -1 
      THEN LEN(CONVERT(NVARCHAR(MAX), qt.text)) * 2 
      ELSE qs.statement_end_offset 
      END - qs.statement_start_offset)/2) + 1) AS [Individual Query] 
  , qt.text AS [Parent Query] 
  , DB_NAME(qt.dbid) AS DatabaseName 
  , qp.query_plan 
FROM sys.dm_exec_query_stats qs 
CROSS APPLY sys.dm_exec_sql_text(qs.sql_handle) as qt 
CROSS APPLY sys.dm_exec_query_plan(qs.plan_handle) qp 
WHERE qs.total_elapsed_time > 0 
ORDER BY [Total CPU time (s)] DESC         

 

原理同上,就不上图了

 

最占IO的前20个查询以及它们的执行计划

 

SET TRANSACTION ISOLATION LEVEL READ UNCOMMITTED 
SELECT TOP 20 
  [Total IO] = (qs.total_logical_reads + qs.total_logical_writes) 
  , [Average IO] = (qs.total_logical_reads + qs.total_logical_writes) / 
                                            qs.execution_count 
  , qs.execution_count 
  , SUBSTRING (qt.text,(qs.statement_start_offset/2) + 1,      
  ((CASE WHEN qs.statement_end_offset = -1 
    THEN LEN(CONVERT(NVARCHAR(MAX), qt.text)) * 2 
    ELSE qs.statement_end_offset 
    END - qs.statement_start_offset)/2) + 1) AS [Individual Query] 
  , qt.text AS [Parent Query] 
  , DB_NAME(qt.dbid) AS DatabaseName 
  , qp.query_plan 
FROM sys.dm_exec_query_stats qs 
CROSS APPLY sys.dm_exec_sql_text(qs.sql_handle) as qt 
CROSS APPLY sys.dm_exec_query_plan(qs.plan_handle) qp 
ORDER BY [Total IO] DESC                                   

结果如图:

5

 

能帮助找出占IO的查询

 

查找被执行次数最多的查询以及它们的执行计划

 

SET TRANSACTION ISOLATION LEVEL READ UNCOMMITTED 
SELECT TOP 20 
    qs.execution_count 
    , SUBSTRING (qt.text,(qs.statement_start_offset/2) + 1,   
    ((CASE WHEN qs.statement_end_offset = -1 
      THEN LEN(CONVERT(NVARCHAR(MAX), qt.text)) * 2 
      ELSE qs.statement_end_offset 
      END - qs.statement_start_offset)/2) + 1) AS [Individual Query] 
    , qt.text AS [Parent Query] 
    , DB_NAME(qt.dbid) AS DatabaseName 
    , qp.query_plan 
FROM sys.dm_exec_query_stats qs 
CROSS APPLY sys.dm_exec_sql_text(qs.sql_handle) as qt 
CROSS APPLY sys.dm_exec_query_plan(qs.plan_handle) qp 
ORDER BY qs.execution_count DESC;    

 

结果如图:

6

 

可以针对用的最多的查询语句做特定优化。

 

特定语句的最后运行时间

 

SET TRANSACTION ISOLATION LEVEL READ UNCOMMITTED 
SELECT DISTINCT TOP 20 
    qs.last_execution_time 
    , qt.text AS [Parent Query] 
    , DB_NAME(qt.dbid) AS DatabaseName 
FROM sys.dm_exec_query_stats qs                       
CROSS APPLY sys.dm_exec_sql_text(qs.sql_handle) as qt 
WHERE qt.text LIKE '%特定语句的部分%' 
ORDER BY qs.last_execution_time DESC            

 

结果如图:

7

posted @ 2015-05-29 11:30  邑尘  阅读(204)  评论(0编辑  收藏  举报