SQL Server 2005부터 추가된 DMV를 통해 여러 가지 유용한 스크립트를 만들 수 있습니다.
이번에는 특정 DB 또는 모든 DB에 대해 SELECT, DELETE, INSERT 또는 UPDATE Statmemt가
어떠한 비율로 발생하는지 그리고 리소스별 어떠한 비율로 사용하는지 쉽게 확인 할 수 있는 쿼리 입니다.
지금까지 막연하게 "이 DB는 대부분이 SELECT가 발생하는 것이야."
이러한 생각을 구체적이고 정확하게 보여줄 수 있습니다.
처음 접하는 시스템을 분석시 위와 같은 자료가 있으면 보다 쉽게 파악 할 수 있을 것입니다.
제 생각에는 ~ 꽤 DB관련 업무에 유용하게 사용될 수 있을 것 같습니다.ㅎㅎ
아래는 수행한 결과 입니다.
ExecutionCount는 분당 실행한 횟수 입니다.
[로그성 DB시스템]
[일반 OLTP DB시스템]
|
with XMLNAMESPACES ('http://schemas.microsoft.com/sqlserver/2004/07/showplan' as sql) SELECT StatementType ,ExecutionCount ,cast((ExecutionCount*1.0/total_ExecutionCount)*100 as Decimal(18,2)) as [% ExecutionCount] ,cast((Elapsed_Time/total_Elapsed_Time)*100 as Decimal(18,2)) as [% Elapsed_Time] ,cast((Worker_Time/total_Worker_Time)*100 as Decimal(18,2)) as [% Worker_Time] ,cast((logical_reads/total_logical_reads)*100 as Decimal(18,2)) as [% logical_reads] ,cast((logical_writes/total_logical_writes)*100 as Decimal(18,2)) as [% logical_writes] ,cast((physical_reads/total_physical_reads)*100 as Decimal(18,2)) as [% physical_reads] FROM ( SELECT StatementType ,sum(avg_ExecutionCount) over(partition by StatementType) as ExecutionCount ,sum(avg_ExecutionCount) over(partition by 1) as total_ExecutionCount ,sum(avg_total_Elapsed_Time) over(partition by StatementType) as Elapsed_Time ,sum(avg_total_Elapsed_Time) over(partition by 1) as total_Elapsed_Time ,sum(avg_total_Worker_Time) over(partition by StatementType) as Worker_Time ,sum(avg_total_Worker_Time) over(partition by 1) as total_Worker_Time ,sum(avg_total_logical_reads) over(partition by StatementType) as logical_reads ,sum(avg_total_logical_reads) over(partition by 1) as total_logical_reads ,sum(avg_total_logical_writes) over(partition by StatementType) as logical_writes ,sum(avg_total_logical_writes) over(partition by 1) as total_logical_writes ,sum(avg_total_physical_reads) over(partition by StatementType) as physical_reads ,case when sum(avg_total_physical_reads) over(partition by 1) = 0 THEN NULL ELSE sum(avg_total_physical_reads) over(partition by 1) END as total_physical_reads FROM( SELECT StatementType ,count(*) as cnt ,sum(ISNULL(ExecutionCount/DATEDIFF(MINUTE , creation_time,GetDate()),0)) as avg_ExecutionCount ,cast(sum(ISNULL([total_Elapsed_Time(sec)]/DATEDIFF(MINUTE , creation_time, GetDate()), 0)) as Decimal(18,2)) as avg_total_Elapsed_Time ,cast(sum(ISNULL([total_Worker_Time(ms)]/DATEDIFF(MINUTE , creation_time, GetDate()), 0)) as Decimal(18,2)) as avg_total_Worker_Time ,cast(sum(ISNULL([total_logical_reads]/DATEDIFF(MINUTE , creation_time, GetDate()), 0)) as Decimal(18,2)) as avg_total_logical_reads ,cast(sum(ISNULL([total_logical_writes]/DATEDIFF(MINUTE , creation_time, GetDate()), 0)) as Decimal(18,2)) as avg_total_logical_writes ,cast(sum(ISNULL(total_physical_reads/DATEDIFF(MINUTE , creation_time, GetDate()), 0)) as Decimal(18,2)) as avg_total_physical_reads FROM ( SELECT db_name ,ISNULL([StatementType],StatementType2) as StatementType --,[statement_text] ,[creation_time] ,[ExecutionCount] ,[total_Elapsed_Time(sec)] ,[total_Worker_Time(ms)] ,[total_logical_reads] ,[total_logical_writes] ,[total_physical_reads] FROM ( SELECT db_name(qp.dbid) AS 'db_name' , qs.creation_time , qs.execution_count AS 'ExecutionCount' , qs.total_elapsed_time/1000.0/1000.0 AS 'total_Elapsed_Time(sec)' , qs.total_worker_time/1000.0 AS 'total_Worker_Time(ms)' , total_logical_reads , qs.total_logical_writes , total_physical_reads , cast(qp.query_plan as xml) as query_plan FROM sys.dm_exec_query_stats AS qs CROSS APPLY sys.dm_exec_text_query_plan(qs.plan_handle,qs.statement_start_offset,qs.statement_end_offset) qp )X OUTER APPLY ( SELECT c.value('(//sql:StmtSimple/@StatementType)[1]','VARCHAR(100)') AS "StatementType" ,c.value('(//sql:StmtCond/@StatementType)[1]','VARCHAR(100)') AS "StatementType2" FROM query_plan.nodes('//sql:QueryPlan')B(C) )xp WHERE db_id(db_name) > 4 AND DATEDIFF(MINUTE,creation_time, GetDate()) > 0 AND StatementType is not null )X group by StatementType )Y )Z |
송 혁, SQL Server MVP
sqler.pe.kr // sqlleader.com
hyoksong.tistory.com
