Oracle分析queryplan

清空Shared pool:alter system flush shared_pool;但为了整个内存的清空，只好将整个Oracle RAC环境的实例和数据库都关机，再重新

清空Shared pool:

alter system flush shared_pool;

但为了整个内存的清空，只好将整个Oracle RAC环境的实例和数据库都关机，再重新启动。

可以采用清空buffer cache的方式来做：

alter system flush buffer_cache;

分析query plan,

explain plan for select * fromHZCZRK_JBXXB,HZCZRK_ZPXXB WHERE HZCZRK_JBXXB.RYID = HZCZRK_ZPXXB.RYID;

select plan_table_output fromtable(dbms_xplan.display());

得到结果：

| Id | Operation |Name | Rows | Bytes |TempSpc| Cost (%CPU)| Time |

-----------------------------------------------------------------------------------------------------------

| 0| SELECT STATEMENT | | 2898K| 572M| | 240K (1)| 00:48:06 |

| 1| MERGE JOIN | | 2898K| 572M| | 240K (1)| 00:48:06 |

| 2| TABLE ACCESS BY INDEX ROWID|HZCZRK_ZPXXB | 2898K| 146M| | 120K (1)| 00:24:07 |

| 3| INDEX FULL SCAN | INDEX_HZCZRK_ZPXXB | 2898K| | | 6771 (1)| 00:01:22 |

|* 4| SORT JOIN | | 2898K| 425M| 2156M| 119K (1)| 00:24:00 |

| 5| TABLE ACCESS FULL | HZCZRK_JBXXB | 2898K| 425M| | 21194 (3)| 00:04:15 |

-----------------------------------------------------------------------------------------------------------

Predicate Information (identified byoperation id):

---------------------------------------------------

4-access("HZCZRK_JBXXB"."RYID"="HZCZRK_ZPXXB"."RYID")

filter("HZCZRK_JBXXB"."RYID"="HZCZRK_ZPXXB"."RYID")

使用set autotrace traceonly后可以查看执行计划、统计信息，，如query:

SELECT COUNT(*) FROM HZCZRK_JBXXB;,执行计划、统计信息如下：

------------------------------------------------------------------------------

| Id | Operation |Name | Rows | Cost (%CPU)| Time |

------------------------------------------------------------------------------

| 0| SELECT STATEMENT | | 1 | 9643 (2)| 00:01:56 |

| 1| SORT AGGREGATE | | 1 | | |

| 2| INDEX FAST FULL SCAN| SYS_C0010867| 17M| 9643 (2)| 00:01:56 |

------------------------------------------------------------------------------

统计信息

----------------------------------------------------------

1 recursive calls

0 db block gets

37509 consistent gets

37485 physical reads

0 redo size

537 bytes sent via SQL*Net toclient

524 bytes received via SQL*Netfrom client

2 SQL*Net roundtrips to/fromclient

0 sorts (memory)

0 sorts (disk)

1 rows processed

consistent gets代表内存消耗，physicalreads代表磁盘I/O消耗，单位都是数据块(DB_BLOCK_SIZE)