提高性能

Question

我有两组来自外部来源的数据 - 客户的购买日期和客户的上次电子邮件点击/开放日期.它分别存储在两个表PURCHASE_INTER和ACTIVITY_INTER表中.购买数据是多个,我需要选择上次购买日期.但活动数据对每个客户都是独一无二的.数据彼此独立,并且可能不存在其他数据集.我们在下面写了一个查询,它结合了两个表,根据person_id对它们进行分组,这是来自外部来源的客户的ID并获取最新的日期,加入我们的客户表以获取客户电子邮件,然后再次加入另一个表最终存储此数据的目的是为了知道它是否是插入或更新操作.您能否建议我如何提高此查询的性能.它非常慢,耗时超过10小时.PURCHASE_INTER和ACTIVITY_INTER表中有数百万条记录.

SELECT INTER.*, C.ID AS CUSTOMER_ID, C.EMAIL AS CUSTOMER_EMAIL, LSI.ID AS INTERACTION_ID, ROW_NUMBER() OVER (ORDER BY PERSON_ID ASC) AS RN FROM (
   SELECT PERSON_ID               AS PERSON_ID,
        MAX(LAST_CLICK_DATE)    AS LAST_CLICK_DATE,
        MAX(LAST_OPEN_DATE)     AS LAST_OPEN_DATE,
        MAX(LAST_PURCHASE_DATE) AS LAST_PURCHASE_DATE
   FROM (
     SELECT ACT.PERSON_ID AS PERSON_ID,
          ACT.LAST_CLICK_DATE AS LAST_CLICK_DATE,
          ACT.LAST_OPEN_DATE AS LAST_OPEN_DATE,
          NULL AS LAST_PURCHASE_DATE
     FROM ACTIVITY_INTER ACT
     WHERE ACT.JOB_ID = 77318317
     UNION
     SELECT PUR.PERSON_ID AS PERSON_ID,
          NULL AS LAST_CLICK_DATE,
          NULL AS LAST_OPEN_DATE,
          PUR.LAST_PURCHASE_DATE AS LAST_PURCHASE_DATE
     FROM PURCHASE_INTER PUR
     WHERE PUR.JOB_ID = 77318317
   ) GROUP BY PERSON_ID
 ) INTER LEFT JOIN CUSTOMER C ON INTER.PERSON_ID = C.PERSON_ID
         LEFT JOIN INTERACTION LSI ON C.ID = LSI.CUSTOMER_ID;

Answer 1

您的查询建议以下索引:

• ACTIVITY_INTER(JOB_ID, PERSON_ID, LAST_CLICK_DATE, LAST_OPEN_DATE)
• PURCHASE_INTER(JOB_ID, PERSON_ID, LAST_PURCHASE_DATE)
• CUSTOMER(PERSON_ID)
• INTERACTION(CUSTOMER_ID)

(对于前两个索引,第一列比其他两个索引更重要,除非匹配数非常大.)

另外,更改UNION为UNION ALL. UNION导致删除重复项的开销 - 这是不可能的(至少在两个子查询之间),因为每个子查询返回不同的列.

此外,您可能希望将第一个子查询替换为full outer join:

SELECT COALESCE(a.PERSON_ID, p.PERSON_ID) as PERSON_ID,
       a.LAST_CLICK_DATE, a.LAST_OPEN_DATE,p.LAST_PURCHASE_DATE
FROM (SELECT ACT.PERSON_ID AS PERSON_ID,
             MAX(ACT.LAST_CLICK_DATE) AS LAST_CLICK_DATE,
             MAX(ACT.LAST_OPEN_DATE) AS LAST_OPEN_DATE
      FROM ACTIVITY_INTER ACT
      WHERE ACT.JOB_ID = 77318317
      GROUP BY ACT.PERSON_ID
     ) a FULL OUTER JOIN
     (SELECT PUR.PERSON_ID AS PERSON_ID,
             MAX(PUR.LAST_PURCHASE_DATE) AS LAST_PURCHASE_DATE
      FROM PURCHASE_INTER PUR
      WHERE PUR.JOB_ID = 77318317
      GROUP BY PER.PERSON_ID
     ) p
     ON a.PERSON_ID = p.PERSON_ID

这为Oracle提供了更多优化选项,因为聚合是直接在表上完成的 - 使索引和更好的统计信息可用于处理.