SQL優化：緊急狀況下提升SQL性能竟是這樣實現的！

在某運營商的優化經歷中曾經遇到了一條比較有意思的 SQL，具體以下：

1 該最開始的 sql 執行狀況以下

SQL> SELECT
  2    NVL(T.RELA_OFFER_SPEC_ID, SUBOS.SUB_OFFER_SPEC_ID) "offerSpecId"
  3    FROM OFFER_SPEC_RELA T
  4    LEFT JOIN OFFER_SPEC_GRP_RELA SUBOS
  5    ON T.RELA_GRP_ID     = SUBOS.OFFER_SPEC_GRP_ID
  6    AND subos.start_dt  <= SYSDATE
  7    AND subos.end_dt    >= SYSDATE
  8    WHERE T.RELA_TYPE_CD = 2
  9    AND t.start_dt      <= SYSDATE
 10    AND t.end_dt        >= SYSDATE
 11    AND (T.OFFER_SPEC_ID = 109910000618
 12    OR EXISTS
 13      (SELECT A.OFFER_SPEC_GRP_ID
 14      FROM OFFER_SPEC_GRP_RELA A
 15      WHERE A.SUB_OFFER_SPEC_ID = 109910000618
 16      AND T.OFFER_SPEC_GRP_ID   = A.OFFER_SPEC_GRP_ID
 17      ))
 18    AND rownum<500;

no rows selected
 
Execution Plan
----------------------------------------------------------
Plan hash value: 1350156609

 

Predicate Information (identified by operation id):
---------------------------------------------------
   1 - filter(ROWNUM<500)
   2 - filter("T"."OFFER_SPEC_ID"=109910000618 OR  EXISTS (SELECT 0 FROM
              "SPEC"."OFFER_SPEC_GRP_RELA" "A" WHERE "A"."OFFER_SPEC_GRP_ID"=:B1 AND
              "A"."SUB_OFFER_SPEC_ID"=109910000618))
   3 - access("T"."RELA_GRP_ID"="SUBOS"."OFFER_SPEC_GRP_ID"(+))
   4 - filter("T"."RELA_TYPE_CD"=2 AND "T"."END_DT">=SYSDATE@! AND
              "T"."START_DT"<=SYSDATE@!)
   5 - filter("SUBOS"."END_DT"(+)>=SYSDATE@! AND "SUBOS"."START_DT"(+)<=SYSDATE@!)
   6 - access("A"."SUB_OFFER_SPEC_ID"=109910000618 AND "A"."OFFER_SPEC_GRP_ID"=:B1)
 
Statistics
----------------------------------------------------------
          0  recursive calls
          0  db block gets
      12444  consistent gets
          0  physical reads
          0  redo size
        339  bytes sent via SQL*Net to client
        509  bytes received via SQL*Net from client
          1  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          0  rows processed
 
                  PLAN                     GET     DISK    WRITE              ROWS      ROWS USER_IO(MS)  ELA(MS)  CPU(MS) CLUSTER(MS)    PLSQL
END_TI I    HASH VALUE EXEC           PRE EXEC PRE EXEC PER EXEC ROW_P    PRE EXEC PRE FETCH    PER EXEC PRE EXEC PRE EXEC    PER EXEC PER EXEC

 

2 第一次分析
此時應該有如下個地方值得注意
1) 該 sql 天天執行上千次，平均每次執行返回不到 10 行數據，可是平均邏輯讀達到1.2W，可能存在性能問題。