[20151014]关于result cache.txt

--11G 开始支持result cache,把执行的结果保存在共享池，从而一定程度减少逻辑读，而那些对象或者那些语句适合做result cache呢？
--自己在前一段时间想把一些表设置RESULT_CACHE为MODE FORCE，访问这些对象时，可以利用result cache模式。
--另外我本来想一些语句通过sql patch的方式加入提示/*+ result_cache */ ,结果不成功。
--参考链接：http://blog.itpub.net/267265/viewspace-1377511/

--那些对象可以这样呢？
--我个人觉得那些执行频次很高的sql语句以及"只读"表(指很少修改的表)做为首选。测试看看：

1.测试环境：
SCOTT@test> @ver1
PORT_STRING          VERSION        BANNER
-------------------- -------------- --------------------------------------------------------------------------------
x86_64/Linux 2.4.xx  11.2.0.3.0     Oracle Database 11g Enterprise Edition Release 11.2.0.3.0 - 64bit Production

SCOTT@test> create table t as select rownum id , lpad('x',200,'x') name  from dual connect by level<=1e4;
Table created.

create unique index pk_t on t (id);
alter table t add constraint pk_t primary key (id);

--分析表忽略。
SCOTT@test> select rowid,t.id from t where id in (1,1000);
ROWID                      ID
------------------ ----------
AABNepAAEAAAACjAAA          1
AABNepAAEAAAAJCAAN       1000

--这两条记录在不同的块中。
--设置result_cache=mode force.

SCOTT@test> alter table t result_cache (mode force);
Table altered.

2.测试：
--注：为了测试方便，打开2个会话，一个设置了set autot traceonly stat;另外一个用来查询相关视图以及一些dml语句。
--session 2：
SCOTT@test> variable x number ;
SCOTT@test> exec :x :=1;
PL/SQL procedure successfully completed.

SCOTT@test> set autot traceonly stat;
SCOTT@test> select * from t where id=:x;
Statistics
----------------------------------------------------------
         34  recursive calls
          0  db block gets
         34  consistent gets
          0  physical reads
          0  redo size
        787  bytes sent via SQL*Net to client
        519  bytes received via SQL*Net from client
          2  SQL*Net roundtrips to/from client
          6  sorts (memory)
          0  sorts (disk)
          1  rows processed

SCOTT@test> select * from t where id=:x;
Statistics
----------------------------------------------------------
          0  recursive calls
          0  db block gets
          0  consistent gets
          0  physical reads
          0  redo size
        787  bytes sent via SQL*Net to client
        519  bytes received via SQL*Net from client
          2  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          1  rows processed

--可以发现第2次执行逻辑读已经为0.说明使用result cache.

3.检查相关视图：

--session 1：
SCOTT@test> SELECT id,TYPE,status,NAME,object_no,cache_id,invalidations,scn FROM v$result_cache_objects;
ID TYPE        STATUS    NAME                              OBJECT_NO CACHE_ID                    INVALIDATIONS          SCN
--- ----------- --------- ------------------------------ ------------ --------------------------- ------------- ------------
  0 Dependency  Published SCOTT.T                              317353 SCOTT.T                                 0  13205305992
  1 Result      Published select * from t where id=:x               0 cxm3q1hv157ps2t68sgc4wp3cg              0  13205305992

SCOTT@test> select * from GV$RESULT_CACHE_DEPENDENCY;
     INST_ID    RESULT_ID    DEPEND_ID    OBJECT_NO
------------ ------------ ------------ ------------
           1            1            0       317353

SCOTT@test> select owner,object_name,object_id,data_object_id from dba_objects where owner=user and object_name in ('T','PK_T');
OWNER  OBJECT_NAME             OBJECT_ID DATA_OBJECT_ID
------ -------------------- ------------ --------------
SCOTT  PK_T                       317354         317354
SCOTT  T                          317353         317353

--可以发现依赖的对象的OBJECT_NO=317353,也就是表T。而不是索引(我以为也会依赖索引)。

4.现在修改id=1000;
--session 1：
SCOTT@test> update t set name=name where id=1000;
1 row updated.

SCOTT@test> commit ;
Commit complete.

--session 2：
SCOTT@test> select * from t where id=:x;
Statistics
----------------------------------------------------------
          0  recursive calls
          0  db block gets
          3  consistent gets
          0  physical reads
          0  redo size
        787  bytes sent via SQL*Net to client
        519  bytes received via SQL*Net from client
          2  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          1  rows processed

SCOTT@test> select * from t where id=:x;
Statistics
----------------------------------------------------------
          0  recursive calls
          0  db block gets
          0  consistent gets
          0  physical reads
          0  redo size
        787  bytes sent via SQL*Net to client
        519  bytes received via SQL*Net from client
          2  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          1  rows processed

--可以发现存在逻辑读，这从一个侧面说明如果这个对象存在修改，会直接导致result cache失效。依赖的对象是表，而不是数据块(感
--觉如果这样代价有点高)。

5.继续测试：
--session 2：
SCOTT@test> select count(*) from t ;
Statistics
----------------------------------------------------------
          1  recursive calls
          0  db block gets
         25  consistent gets
          0  physical reads
          0  redo size
        526  bytes sent via SQL*Net to client
        519  bytes received via SQL*Net from client
          2  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          1  rows processed

SCOTT@test> select count(*) from t ;
Statistics
----------------------------------------------------------
          0  recursive calls
          0  db block gets
          0  consistent gets
          0  physical reads
          0  redo size
        526  bytes sent via SQL*Net to client
        519  bytes received via SQL*Net from client
          2  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          1  rows processed
--这个语句执行计划应该使用主键索引。

--session 1：
SCOTT@test> SELECT id,TYPE,status,NAME,object_no,cache_id,invalidations,scn FROM v$result_cache_objects;
ID TYPE        STATUS    NAME                              OBJECT_NO CACHE_ID                    INVALIDATIONS          SCN
--- ----------- --------- ------------------------------ ------------ --------------------------- ------------- ------------
  0 Dependency  Published SCOTT.T                              317353 SCOTT.T                                 1  13205306169
  3 Result      Published select count(*) from t                    0 gp1k5qq6gk8spd439xt91vnwb0              0  13205306238
  2 Result      Published select * from t where id=:x               0 cxm3q1hv157ps2t68sgc4wp3cg              0  13205306171
  1 Result      Invalid   select * from t where id=:x               0 cxm3q1hv157ps2t68sgc4wp3cg              0  13205305992

SCOTT@test> select * from GV$RESULT_CACHE_DEPENDENCY;
     INST_ID    RESULT_ID    DEPEND_ID    OBJECT_NO
------------ ------------ ------------ ------------
           1            3            0       317353
           1            2            0       317353

--即使执行select count(*) from t ;，我开始以为这样会依赖主键索引，结果依赖的还是表T。
--说明一下：RESULT_ID=3的记录，对应DEPEND_ID=0,而SELECT * FROM v$result_cache_objects where DEPEND_ID=0的记录表scott.T .
--这样说明即使存在DML语句也会导致result cache失效。

6.修改记录看看：
-- session 1:
SCOTT@test> select * from  t where id=999 for update ;
ID NAME
--- -----
999 test

SCOTT@test> commit ;
Commit complete.

-- session 2:
SCOTT@test> select count(*) from t ;
Statistics
----------------------------------------------------------
          0  recursive calls
          0  db block gets
         25  consistent gets
          0  physical reads
          0  redo size
        526  bytes sent via SQL*Net to client
        519  bytes received via SQL*Net from client
          2  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          1  rows processed

--说明：我选择select .. for update 测试，结果一样。result cache失效。

SCOTT@test> SELECT id,TYPE,status,NAME,object_no,cache_id,invalidations,scn FROM v$result_cache_objects;
ID TYPE        STATUS    NAME                              OBJECT_NO CACHE_ID                    INVALIDATIONS          SCN
--- ----------- --------- ------------------------------ ------------ --------------------------- ------------- ------------
  0 Dependency  Published SCOTT.T                              317353 SCOTT.T                                 3  13205306490
  5 Result      Published select count(*) from t                    0 gp1k5qq6gk8spd439xt91vnwb0              0  13205306493
  1 Result      Invalid   select * from t where id=:x               0 cxm3q1hv157ps2t68sgc4wp3cg              0  13205305992
  2 Result      Invalid   select * from t where id=:x               0 cxm3q1hv157ps2t68sgc4wp3cg              0  13205306171
  3 Result      Invalid   select count(*) from t                    0 gp1k5qq6gk8spd439xt91vnwb0              0  13205306238
  4 Result      Invalid   select count(*) from t                    0 gp1k5qq6gk8spd439xt91vnwb0              0  13205306433

6 rows selected.

--我感觉依赖的本质是scn。继续测试就可以看到问题了：

7.修改记录然后回滚。
SCOTT@test> update t set name=name where id=998;
1 row updated.

SCOTT@test> rollback;
Rollback complete.

SCOTT@test> select count(*) from t ;
Statistics
----------------------------------------------------------
          0  recursive calls
          0  db block gets
          0  consistent gets
          0  physical reads
          0  redo size
        526  bytes sent via SQL*Net to client
        519  bytes received via SQL*Net from client
          2  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          1  rows processed
-- result cache 有效!

SCOTT@test> update t set name=name where id=998;
1 row updated.
--不提交

SCOTT@test> select count(*) from t ;
Statistics
----------------------------------------------------------
          0  recursive calls
          0  db block gets
          0  consistent gets
          0  physical reads
          0  redo size
        526  bytes sent via SQL*Net to client
        519  bytes received via SQL*Net from client
          2  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          1  rows processed

--我没有提交或者执行了rollback，都不会影响结果集。

SCOTT@test> commit ;
Commit complete.

SCOTT@test> select count(*) from t ;

Statistics
----------------------------------------------------------
          0  recursive calls
          0  db block gets
         25  consistent gets
          0  physical reads
          0  redo size
        526  bytes sent via SQL*Net to client
        519  bytes received via SQL*Net from client
          2  SQL*Net roundtrips to/from client
          0  sorts (memory)
          0  sorts (disk)
          1  rows processed

--consistent gets=25。结果集失效。

8.总结：
---从以上测试，可以说明result cache 使用范围仅仅合适频繁执行，涉及对象表少量修改的语句。或者也许一些数据仓库DML结果集很少变
--化的情况。这样才能一定程度减少逻辑读，提高性能。

--另外可以通过查询视图确定v$result_cache_objects,确定那些对象不合适做result cache。
--看看生产系统：

> SELECT id,TYPE,status,NAME,object_no,cache_id,invalidations,scn FROM v$result_cache_objects where INVALIDATIONS>=1000;
     ID TYPE        STATUS    NAME                    OBJECT_NO CACHE_ID            INVALIDATIONS          SCN
------- ----------- --------- -------------------- ------------ ------------------- ------------- ------------
336708 Dependency  Published XXXXXX_XXX.YF_DB01          96396 XXXXXX_XXX.YF_DB01           8420  13860014605
  39171 Dependency  Published XXXXXX_XXX.GY_XTCS          94089 XXXXXX_XXX.GY_XTCS         318486  13860074094
  39283 Dependency  Published XXXXXX_XXX.GY_YHCS          94111 XXXXXX_XXX.GY_YHCS         499397  13860074730
      3 Dependency  Published XXXXXX_XXX.GY_YGDM          94105 XXXXXX_XXX.GY_YGDM           4094  13859061872

--这些对象也许不适合修改表的result cache属性为force。当然要仔细评估。我们的系统刷屏语句实在太厉害了。

--我简单写一个脚本：
$ cat d_buffer.sql
col executions1 new_value x1
col buffer_gets1 new_value x2
col ELAPSED_TIME1 new_value x3
col ROWS_PROCESSED1 new_value x4

col executions2 new_value y1
col buffer_gets2 new_value y2
col ELAPSED_TIME2 new_value y3
col ROWS_PROCESSED2 new_value y4

select executions executions1,buffer_gets buffer_gets1,elapsed_time elapsed_time1,rows_processed rows_processed1  from v$sqlarea where sql_id='&&1';
prompt ... sleep &&2 , waiting ....
host sleep &&2
select executions executions2,buffer_gets buffer_gets2,elapsed_time elapsed_time2 ,rows_processed rows_processed2 from v$sqlarea where sql_id='&&1';
select (&y2-&x2)/(&y1-&x1) "每次buffer_gets",&y1- &x1 "执行次数" ,  (&y3-&x3)/(&y1-&x1) "每次执行时间" , (&y4-&x4)/(&y1-&x1) "平均处理记录数" from dual ;

SQL> select sql_id,sql_text,executions from v$sqlarea where sql_id='ff511u9ndctfr';
SQL_ID        SQL_TEXT                                                       EXECUTIONS
------------- ------------------------------------------------------------ ------------
ff511u9ndctfr Select CSZ From GY_XTCS Where CSMC =:1                             250432

SQL> @d_buffer ff511u9ndctfr 10

EXECUTIONS1 BUFFER_GETS1 ELAPSED_TIME1 ROWS_PROCESSED1
------------ ------------ ------------- ---------------
      255148       200085      15952863          255150

... sleep 10 , waiting ....

EXECUTIONS2 BUFFER_GETS2 ELAPSED_TIME2 ROWS_PROCESSED2
------------ ------------ ------------- ---------------
      255657       200397      15980172          255659

每次buffer_gets     执行次数 每次执行时间 平均处理记录数
--------------- ------------ ------------ --------------
   .61296660118          509 53.652259332              1

--10内执行509次，平均下来50次/秒。真不知道开发什么写代码的？？？？