*任何人都可以帮我优化查询。我在一个名为streamsets的ETL中使用此查询，当我运行一个非常慢的streamsets pipleline时，它在6分钟内产生70条记录。我们从SSIS包中获取此查询并使用左外连接连接每个表。我需要对其进行优化，以便在streamsets *
中每分钟获取至少1000条记录 SELECT [FR].[ORDER_ID]
, [FR].[FULFILLMENT_REQUEST_ID]
, [SR].[SCREENING_RESULT_ID]
, [AT].[AG_TASK_ID]
, [RCC].[RESULT_CRIM_CASE_ID]
, [RCC].[RESULT_CRIM_CHARGE_ID]
, [FR].[JURISDICTION_ID]
, [COUNTY] = CASE WHEN [FR].[PRODUCT_ID] = 2 THEN [JC].[COUNTY_NAME] ELSE '' END
, [STATE] = CASE WHEN [FR].[PRODUCT_ID] = 2 THEN [JC].[STATE_A2_CD] ELSE [JS].[STATE_A2_CD] END
, [JT].[JURISDICTION_TYPE_NAME]
, [FR].[PRODUCT_ID]
, [P].[PRODUCT_NAME]
, [FR].[ORGANIZATION_ID]
, [O].[ORGANIZATION_NAME]
, [SA].[SALARY_RANGE]
, [AGE] = YEAR(GETDATE()) - YEAR([FR].[SUBJECT_DOB])
, [FR].[SUBJECT_JOB_COUNTRY]
, [FR].[SUBJECT_JOB_STATE]
, [FR].[ENTRY_DATE]
, [FR].[CREATION_DATE]
, [FR].[CLOSED_DATE]
, [FR].[EXTERNAL_USER_ID]
, [FR].[GENDER_CODE_ID]
, [GC].[GENDER_CODE_VALUE]
, [AT].[DATA_SOURCE]
, [SU].[RESEARCHER_CLASS]
, [TT].[TASK_TYPE]
, [TT].[TASK_DESCRIPTION]
, [RCC].[CHARGE_DESCRIPTION]
, [RCC].[DISPOSITION_DESCRIPTION]
, [P].[GENERAL_LEDGER_NBR]
, [O].[SCHEME_AGENCY_ID]
, [RT].[RESULT_TYPE_ID]
, [RT].[RESULT_CATEGORY]
, [RT].[RESULT_TYPE]
, [SU].[USER_ID]
,FR.AG_STATUS_ID
, [SCREENING_REFERENCE_ID]=CAST(LEFT([FR].[SCREENING_REFERENCE_ID],250) AS VARCHAR(8000))
,[P1].POD_NAME
, ATH.AG_TASK_HISTORY_ID
,ATH.TASK_STATUS TASK_HISTORY_STATUS
, ATH.AG_TASK_DATE TASK_HISTORY_DATE
,[TT].[TASK_TYPE_ID]
,ATH.MODIFIED_BY_SYSTEM_USER_ID
,ATH.CF_SYSTEM_USER_ID
,'' AS [STATUS NOTES]
, FR.SCOPE_OF_SEARCH
, SR.COMMONNAMFLAG
,CASE WHEN SU.RESEARCHER_CLASS = 'EXTERNAL' THEN SU.[USER_ID] END AS VENDOR
, CASE [TT].TASK_TYPE_ID WHEN 14 THEN 'SD IN' WHEN 15 THEN 'MR' WHEN 1 THEN 'ER' END AS [REQUIRED_ACTION]
, CASE
WHEN [TT].TASK_TYPE_ID = 1 THEN
CASE ATH.TASK_STATUS
WHEN 'DOCUMENTUPLOADED' THEN 'FMT RECORD ENTRY'
WHEN 'ACKNOWLEDGED' THEN 'SEARCHES'
WHEN 'AWAITINGHITENTRY' THEN 'AWAITING HIT'
WHEN 'DISPUTE' THEN 'AWAITING HIT'
WHEN 'DOUBLEENTERRESULTS' THEN 'AWAITING HIT'
WHEN 'NEW' THEN 'SEARCHES'
WHEN 'SOURCEUNAVAILABLE' THEN 'SEARCHES'
WHEN 'UPLOADED' THEN 'SEARCHES'
END
WHEN [TT].TASK_TYPE_ID = 15 THEN
CASE ATH.TASK_STATUS
WHEN 'DISPUTE' THEN 'QC'
WHEN 'DOCUMENTUPLOADED' THEN 'QC'
WHEN 'DOUBLESMARTDATAREVIEW' THEN 'QC'
WHEN 'MANUALREVIEW' THEN 'QC'
END
WHEN [TT].TASK_TYPE_ID = 14 THEN
CASE ATH.TASK_STATUS
WHEN 'DISPUTE' THEN 'QC'
WHEN 'INFONEEDED' THEN 'QC'
WHEN 'INFOPROVIDED' THEN 'QC'
WHEN 'INFONEEDEDACK' THEN 'QC'
WHEN 'NEW' THEN 'QC'
WHEN 'SDINFONEEDED' THEN 'CD COMPLIANCE'
END
END AS [AOM STATUS]
, CASE
WHEN [TT].TASK_TYPE_ID = 1 AND
ATH.TASK_STATUS = 'DOCUMENTUPLOADED' AND
ATH.MODIFIED_BY_SYSTEM_USER_ID IS NOT NULL AND
ATH.CF_SYSTEM_USER_ID IS NOT NULL THEN 'ER DOCUMENT UPLOADED COMPLETE'
WHEN [TT].TASK_TYPE_ID = 14 AND
ATH.TASK_STATUS = 'INFONEEDED' THEN 'SD INFO NEEDED INFO NEEDED DATA'
WHEN [TT].TASK_TYPE_ID = 1 AND SU.RESEARCHER_CLASS = 'INTERNAL' AND
FR.AG_STATUS_ID NOT IN (152) THEN 'ER INTERNAL'
WHEN [TT].TASK_TYPE_ID = 1 AND
ATH.TASK_STATUS = 'AWAITINGHITENTRY' AND
ATH.MODIFIED_BY_SYSTEM_USER_ID IS NOT NULL AND
ATH.CF_SYSTEM_USER_ID IS NOT NULL AND
ATH.MODIFIED_BY_SYSTEM_USER_ID = ATH.CF_SYSTEM_USER_ID THEN 'ER AWAITING HIT ENTRY DATA'
WHEN [TT].TASK_TYPE_ID = 14 THEN 'SD INFO NEEDED DATA'
WHEN [TT].TASK_TYPE_ID = 15 AND
ATH.TASK_STATUS = 'INFONEEDED' THEN 'MR INFO NEEDED'
END AS AG_OPS_STATUS
, FR.EXTERNAL_ORDER_ID
, AOT.OTD
, drv_pod_name=[P1].POD_NAME
, TAT_IN_MIN= Case when TAT_IN_MIN < 0 then 0 else TAT_IN_MIN end
, SU2.USER_ID CF_SYS_USER
, SU3.USER_ID MODIFIED_SYS_USER
, GETDATE() GET_DATE
FROM AG_TASK_HISTORY [ATH] LEFT OUTER JOIN
(
(
(
(
(
(
(
(
(
SELECT [ORDER_ID] ,[FULFILLMENT_REQUEST_ID],[JURISDICTION_ID],[PRODUCT_ID],[ORGANIZATION_ID],[SUBJECT_DOB],[SUBJECT_JOB_COUNTRY],[SUBJECT_JOB_STATE],[ENTRY_DATE],[CREATION_DATE],[CLOSED_DATE],[EXTERNAL_USER_ID],[GENDER_CODE_ID],AG_STATUS_ID,[SCREENING_REFERENCE_ID],SCOPE_OF_SEARCH,EXTERNAL_ORDER_ID,POD_ID,[SALARY_RANGE_ID],[JURISDICTION_TYPE]
FROM [FULFILLMENT_REQUEST]
) FR
LEFT JOIN
(
(SELECT [SCREENING_RESULT_ID],[RESULT_TYPE_ID],[FULFILLMENT_REQUEST_ID],COMMONNAMFLAG FROM [SCREENING_RESULT] )[SR]
INNER JOIN [DBO].[RESULT_TYPE] [RT]
ON [RT].[RESULT_TYPE_ID] = [SR].[RESULT_TYPE_ID]
)
ON [SR].[FULFILLMENT_REQUEST_ID] = [FR].[FULFILLMENT_REQUEST_ID]
)
INNER JOIN [PRODUCT] [P]
ON [P].[PRODUCT_ID] = [FR].[PRODUCT_ID]
INNER JOIN [ORGANIZATION] [O]
ON [O].[ORGANIZATION_ID] = [FR].[ORGANIZATION_ID]
LEFT JOIN(
(SELECT [AG_TASK_ID],[DATA_SOURCE],[TASK_TYPE_ID],[FULFILLMENT_REQUEST_ID],[SYSTEM_USER_ID] FROM [AG_TASK] ) [AT]
INNER JOIN [TASK_TYPE] [TT]
ON [TT].[TASK_TYPE_ID] = [AT].[TASK_TYPE_ID]
)
ON [AT].[FULFILLMENT_REQUEST_ID] = [FR].[FULFILLMENT_REQUEST_ID])
LEFT OUTER JOIN (
(SELECT [RESULT_CRIM_CASE_ID],[SCREENING_RESULT_ID] FROM [RESULT_CRIM_CASE])[RC]
INNER JOIN (SELECT [RESULT_CRIM_CASE_ID],[RESULT_CRIM_CHARGE_ID],[CHARGE_DESCRIPTION],[DISPOSITION_DESCRIPTION] FROM [RESULT_CRIM_CHARGE]) [RCC]
ON [RCC].[RESULT_CRIM_CASE_ID] = [RC].[RESULT_CRIM_CASE_ID]
)
ON [SR].[SCREENING_RESULT_ID] = [RC].[SCREENING_RESULT_ID]
)
)
LEFT JOIN [SYSUSER] [SU]
ON [SU].[SYSTEM_USER_ID] = [AT].[SYSTEM_USER_ID]
LEFT OUTER JOIN [SALARY_RANGE] [SA]
ON [SA].[SALARY_RANGE_ID] = [FR].[SALARY_RANGE_ID]
)
LEFT OUTER JOIN [JURISDICTION_TYPE] [JT]
ON [JT].[JURISDICTION_TYPE_ID] = [FR].[JURISDICTION_TYPE]
)
LEFT OUTER JOIN [GENDER_CODE] [GC]
ON [GC].[GENDER_CODE] = [FR].[GENDER_CODE_ID]
)
LEFT OUTER JOIN [JURISDICTION_COUNTY] [JC]
ON [JC].[JURISDICTION_ID] = [FR].[JURISDICTION_ID]
LEFT OUTER JOIN [JURISDICTION_STATE] [JS]
ON [JS].[JURISDICTION_ID] = [FR].[JURISDICTION_ID]
)
ON [ATH].AG_TASK_ID=[AT].AG_TASK_ID
LEFT OUTER JOIN [POD] [P1] ON [FR].POD_ID = [P1].POD_ID
LEFT OUTER JOIN(
SELECT *, (DATEDIFF(MINUTE, [IN_DATE], [IP_DATE]))-(DATEDIFF(WK, [IN_DATE], [IP_DATE]) * (2*24*60))-
(CASE WHEN DATENAME(DW, [IN_DATE]) = 'SUNDAY'
THEN (24*60) ELSE 0 END)
-(CASE WHEN DATENAME(DW, [IP_DATE]) = 'SATURDAY'
THEN (24*60) ELSE 0 END) TAT_IN_MIN
FROM (
SELECT FULFILLMENT_REQUEST_ID , MAX([IN_DATE]) [IN_DATE], MAX([IP_DATE]) [IP_DATE]
FROM(
SELECT AG_TASK_HISTORY.FULFILLMENT_REQUEST_ID,
CASE WHEN (AG_TASK_HISTORY.TASK_STATUS='INFONEEDED' OR AG_TASK_HISTORY.TASK_STATUS='NEW') AND (SYSUSER.RESEARCHER_CLASS='EXTERNAL' OR SU1.RESEARCHER_CLASS ='EXTERNAL')
THEN AG_TASK_HISTORY.AG_TASK_DATE END [IN_DATE],
CASE WHEN (AG_TASK_HISTORY.TASK_STATUS='INFOPROVIDED' OR AG_TASK_HISTORY.TASK_STATUS='COMPLETE' OR AG_TASK_HISTORY.TASK_STATUS='DOCUMENTUPLOADED') AND (SYSUSER.RESEARCHER_CLASS='EXTERNAL' OR SU1.RESEARCHER_CLASS ='EXTERNAL')
THEN AG_TASK_HISTORY.AG_TASK_DATE END AS [IP_DATE]
FROM AG_TASK_HISTORY
LEFT JOIN SYSUSER(NOLOCK) ON AG_TASK_HISTORY.CF_SYSTEM_USER_ID = SYSUSER.SYSTEM_USER_ID
LEFT JOIN SYSUSER(NOLOCK) SU1 ON AG_TASK_HISTORY.MODIFIED_BY_SYSTEM_USER_ID = SU1.SYSTEM_USER_ID
WHERE AG_TASK_DATE >=CAST('01-JAN-'+CAST(YEAR(GETDATE())-3 AS CHAR(4))AS DATETIME)
) A GROUP BY FULFILLMENT_REQUEST_ID
)B
)AVT
ON FR.FULFILLMENT_REQUEST_ID = AVT.FULFILLMENT_REQUEST_ID
LEFT OUTER JOIN(
SELECT *, (DATEDIFF(DAY, [IN_DATE], [IP_DATE]))-
(DATEDIFF(WK, [IN_DATE], [IP_DATE]) * (2))-
(CASE WHEN DATENAME(DW, [IN_DATE]) = 'SUNDAY' THEN 1 ELSE 0 END)
-(CASE WHEN DATENAME(DW, [IP_DATE]) = 'SATURDAY' THEN 1 ELSE 0 END) OTD
FROM
(
SELECT FULFILLMENT_REQUEST_ID , MIN(CREATION_DATE) [IN_DATE], GETDATE() [IP_DATE]
FROM FULFILLMENT_REQUEST WHERE CLOSED_DATE IS NULL AND AG_STATUS_ID NOT IN (29,134,142,152)
GROUP BY FULFILLMENT_REQUEST_ID
)B
) AOT
ON FR.FULFILLMENT_REQUEST_ID = AOT.FULFILLMENT_REQUEST_ID
LEFT OUTER JOIN SYSUSER SU2
ON ATH.CF_SYSTEM_USER_ID=SU2.SYSTEM_USER_ID
LEFT OUTER JOIN SYSUSER SU3
ON ATH.MODIFIED_BY_SYSTEM_USER_ID=SU3.SYSTEM_USER_ID
WHERE SU.RESEARCHER_CLASS='EXTERNAL' AND (AVT.TAT_IN_MIN >0 OR AOT.[IP_DATE] IS NULL)
AND ATH.AG_Task_Date >=' 2017-07-07 09:02:23.050'
答案 0 :(得分：3)
如果你连接表和子选择，你真的必须知道你的索引。 使用说明检查使用索引是否存在问题。 查询可能无法优化，必须完全重新设计。
答案 1 :(得分：2)
很难给出明确的答案，因为您的查询很大，需要很长时间才能对其进行分析。
以下是一些建议：
其他考虑因素是使用存储过程（如果可能的话）并将这个大查询拆分成更小的部分，但我很确定使用适当的索引集可以获得比现在更好的性能。