查询生产耗时

Question

我们有这样的查询，我们试图识别具有多个信用选项指示器的客户。我们必须在报告中反映此查询输出，并将其分享给业务用户。我们几乎每周都要运行这个查询，而且这个查询需要时间。

查询

select CUST_ID, CUST_COUNT from (
SELECT N.CONS_ID AS CUST_ID,
COUNT(DISTINCT(case when M.CO_ID is null then 1 else m.co_id end)) AS CUST_COUNT
FROM CTS_VIEW.CNSLD_CREDIT_SUM M
INNER JOIN  CTS_VIEW.LEGACY_CODE_XREF  N
ON M.EE_ID = N.EE_GBL_IND
WHERE M.PROD_DT >= DATE '2018-12-31'
GROUP BY N.CONS_ID
  ) a
where CUST_COUNT>1;

有没有更好的方法来编写这个查询来加速执行时间。我们已经应用了CBO，并在会话级别启用了矢量化。

Answer 1

也许使用这样的临时表可以使查询速度更快。同时，您应该检查这些表所需的索引。

SELECT N.CONS_ID AS CUST_ID,
COUNT(Distinct(Isnull(M.CO_ID,1))) AS CUST_COUNT
Into #Temp
FROM CTS_VIEW.CNSLD_CREDIT_SUM M
INNER JOIN  CTS_VIEW.LEGACY_CODE_XREF  N
ON M.EE_ID = N.EE_GBL_IND
WHERE M.PROD_DT >= DATE '2018-12-31'
GROUP BY N.CONS_ID

select CUST_ID, CUST_COUNT from 
#Temp
where CUST_COUNT>1;

Answer 2

我认为CTE或Common Table表达式对于这些目的来说是很好的-

SELECT CUST_ID,CUST_COUNT FROM (
WITH M_RAW_CTE AS 
(SELECT CO_ID,EE_ID,PROD_DT FROM CTS_VIEW.CNSLD_CREDIT_SUM),
M_CTE AS 
(SELECT * FROM M_RAW_CTE WHERE PROD_DT >= DATE '2018-12-31'), 
N_CTE AS 
(SELECT CONS_ID,EE_GBL_IND FROM LEGACY_CODE_XREF) 

SELECT N_CTE.CONSID AS CUST_ID,
COUNT(DISTINCT(ISNULL(M.CO_ID,1))) AS CUST_COUNT
INNER JOIN N_CTE ON M_CTE.EE_ID = N_CTE.EE_GBL_IND
GROUP BY N_CTE.CONS_ID)
WHERE CUST_COUNT > 1;

使用CTE背后的概念是提高DDL中数据的可重用性。