oracle慎用基於on commit重新整理物化檢視的方式（二）
阿新 • • 發佈：2020-08-17
oracle慎用基於on commit重新整理物化檢視的方式（二）
環境模擬
首先，需要有示例使用者SCOTT，以emp表為例子。
首先是序列下無物化檢視的emp表，做update操作的情況，
16:56:14 SYS@zkm(27)> begin 16:56:21 2 for i in 1..10000 loop 16:56:21 3 update scott.emp set sal=9999 where empno=7934; 16:56:21 4 commit; 16:56:21 5 end loop; 16:56:21 6 end ; 16:56:21 7 / PL/SQL procedure successfully completed. Elapsed: 00:00:00.48 16:56:23 SYS@zkm(27)> begin 16:56:26 2 for i in 1..10000 loop 16:56:26 3 update scott.emp set sal=9999 where empno=7934; 16:56:26 4 commit; 16:56:26 5 end loop; 16:56:26 6 end; 16:56:26 7 / PL/SQL procedure successfully completed. Elapsed: 00:00:00.65 16:56:27 SYS@zkm(27)> begin 16:56:28 2 for i in 1..10000 loop 16:56:28 3 update scott.emp set sal=9999 where empno=7934; 16:56:28 4 commit; 16:56:28 5 end loop; 16:56:28 6 end ; 16:56:28 7/ PL/SQL procedure successfully completed. Elapsed: 00:00:00.65
然後是序列下有物化檢視的emp表，做update操作的情況，
17:15:23 SYS@zkm(27)> conn scott/oracle Connected. 17:15:32 SCOTT@zkm(27)> create materialized view scott.mv_emp 17:15:33 2 REFRESH force on commit 17:15:33 3 as 17:15:33 4 select * from scott.emp; Materialized view created. Elapsed: 00:00:00.55 17:15:34 SCOTT@zkm(27)> begin 17:16:06 2 for i in 1..10000 loop 17:16:06 3 update scott.emp set sal=9999 where empno=7934; 17:16:06 4 commit; 17:16:06 5 end loop; 17:16:06 6 end ; 17:16:06 7 / PL/SQL procedure successfully completed. Elapsed: 00:00:21.73 17:16:29 SCOTT@zkm(27)> / PL/SQL procedure successfully completed. Elapsed: 00:00:21.64 17:16:53 SCOTT@zkm(27)> / PL/SQL procedure successfully completed. Elapsed: 00:00:20.70
可以看到，有物化檢視的情況下，效能下降40多倍。
更不要說在並行情況下產生行級鎖的情況了，幾乎會導致業務都癱瘓掉。
並行下也進行一次模擬看看，開啟兩個session，sid分別為27和31。
刪除上邊的物化檢視，
17:33:58 SCOTT@zkm(27)> drop materialized view scott.mv_emp; Materialized view dropped. Elapsed: 00:00:00.50
為了保證同時執行匿名塊，使用crt的command windows功能，將匿名塊同時傳送到活動的會話裡邊，
可以發現，執行時間情況為
17:34:07 SCOTT@zkm(27)> begin 17:34:32 2 for i in 1..10000 loop 17:34:32 3 update scott.emp set sal=9999 where empno=7934; 17:34:32 4 commit; 17:34:32 5 end loop; 17:34:32 6 end ; 17:34:32 7 / PL/SQL procedure successfully completed. Elapsed: 00:00:01.06 17:34:33 SCOTT@zkm(27)> / PL/SQL procedure successfully completed. Elapsed: 00:00:01.12 17:34:39 SCOTT@zkm(27)> / PL/SQL procedure successfully completed. Elapsed: 00:00:01.17 17:34:43 SCOTT@zkm(27)>
----------------------------同時sid=31的執行----------------------------- 17:34:16 SCOTT@zkm(31)> begin 17:34:32 2 for i in 1..10000 loop 17:34:32 3 update scott.emp set sal=9999 where empno=7934; 17:34:32 4 commit; 17:34:32 5 end loop; 17:34:32 6 end ; 17:34:32 7 / PL/SQL procedure successfully completed. Elapsed: 00:00:01.13 17:34:33 SCOTT@zkm(31)> / PL/SQL procedure successfully completed. Elapsed: 00:00:01.18 17:34:39 SCOTT@zkm(31)> / PL/SQL procedure successfully completed. Elapsed: 00:00:01.24 17:34:43 SCOTT@zkm(31)>
相對於序列下，效能稍微下降。
再試試有物化檢視下，併發的執行情況，
17:40:44 SCOTT@zkm(27)> create materialized view scott.mv_emp 17:40:44 2 REFRESH force on commit 17:40:44 3 as 17:40:44 4 select * from scott.emp; Materialized view created. Elapsed: 00:00:00.03
17:40:47 SCOTT@zkm(27)> begin 17:41:01 2 for i in 1..10000 loop 17:41:01 3 update scott.emp set sal=9999 where empno=7934; 17:41:01 4 commit; 17:41:01 5 end loop; 17:41:01 6 end ; 17:41:01 7 / PL/SQL procedure successfully completed. Elapsed: 00:00:44.82 17:41:46 SCOTT@zkm(27)> / PL/SQL procedure successfully completed. Elapsed: 00:00:49.45
----------------------------同時sid=31的執行-----------------------------
17:34:43 SCOTT@zkm(31)> begin 17:41:01 2 for i in 1..10000 loop 17:41:01 3 update scott.emp set sal=9999 where empno=7934; 17:41:01 4 commit; 17:41:01 5 end loop; 17:41:01 6 end ; 17:41:01 7 / PL/SQL procedure successfully completed. Elapsed: 00:00:44.78 17:41:46 SCOTT@zkm(31)> / PL/SQL procedure successfully completed. Elapsed: 00:00:49.32
效能相對序列下降了將近一倍。
如果併發的會話是3個或者更多，那麼預計效能下降會更大，這裡不做測試。