我加了很多ORACLE群，自己本人也很喜欢扯淡。最近经常去ORACLE酱油群扯淡，酱油群里有很多大师，还有ORACLE原厂的，神马世界500强的都有。
经常看见盖尔拿个SQL出来吆喝，我这人有个爱好，喜欢优化复杂SQL，对神马统计信息缺失/不准，绑定变量窥探，等等的SQL优化很没兴趣，也对神马聚簇因子很大滴优化也没兴趣，就喜欢搞复杂的SQL，其实优化复杂SQL往往需要改写代码，但是我这个人懒，又偏偏不喜欢改写SQL，总是喜欢给SQL找点性能瓶颈，让他们自己改去 呵呵。
盖尔的那个SQL信息如下：
select *
from (select rownum as my_rownum, table_a.*
from (select e.company_name organ_name,
e.abbr_name organ_abbr_name,
a.agency_uw_state,
a.agency_uw_time,
a.ht_bill_def_version,
(SELECT version.ht_bill_def_version_type_id
FROM t_ht_bill_def_version version
WHERE a.ht_bill_def_version =
version.ht_bill_def_version) ht_bill_def_version_type_id,
a.policy_id,
tp.p_state_id,
a.organ_id,
a.agent_id,
tp.send_code,
a.policy_code,
a.insert_time,
(SELECT tct.real_name
FROM t_customer tct
WHERE tct.customer_id = a.applicant_id) holder_name,
(SELECT tbe.bank_code
FROM t_bank_employee tbe
WHERE a.agency_hand = tbe.emp_id) bank_code,
a.agency_code,
(SELECT b.is_charge
FROM t_agent b
WHERE a.agent_id = b.agent_id) is_charge,
(SELECT b.real_name
FROM t_agent b
WHERE a.agent_id = b.agent_id) agent_name,
tp.record_end,
tp.inspect_time,
tp.record_insert_time,
a.derivation,
tp.pending_cause,
a.update_time,
tp.scan_time,
tp.first_end,
tp.priority_id,
a.policy_type,
(SELECT f.real_name
FROM t_employee f
WHERE tp.first_user_id = f.emp_id) first_name,
'' recorder_name,
(SELECT g.real_name
FROM t_employee g
WHERE a.agency_uw_rec_id = g.emp_id) checker_name,
pr.apply_time,
pr.finish_time,
pr.handler_id,
pr.apply_id
from t_contract_master a,
t_company_organ e,
t_policy tp,
t_policy_reentry pr
where a.organ_id = e.organ_id
and a.policy_id = tp.policy_id
and pr.policy_id(+) = a.policy_id
and (pr.reentry_id =
(select max(tpr.reentry_id)
from t_policy_reentry tpr
where tpr.policy_id = a.policy_id) or
pr.reentry_id is null)
and pr.reentry_id is null
and a.sell_way in ('2', '3', '12', '18')
and a.agency_uw_state <> '2'
and not EXISTS (SELECT 1
FROM t_agency_prolicy_problem tap
WHERE tap.POLICY_ID = a.POLICY_ID)
and (select count(*)
from t_policy_problem tpp
where tpp.origin_type = 802
and tpp.policy_id = a.policy_id) = 0
and (select count(*)
from t_policy_problem tpp
where tpp.origin_type = 801
and tpp.policy_id = a.policy_id) = 0
and a.sell_way = '3'
and a.sale_channel in ('3', '5')
and EXISTS (SELECT 1
FROM t_image tii
WHERE tii.POLICY_ID = a.POLICY_ID
and tii.image_type_id in (308, 309))
and (select trunc(min(ti.scan_time))
from t_image ti
WHERE ti.image_type_id in (308, 309)
and ti.policy_id = a.policy_id) >=
to_date('2011-01-01', 'YYYY-MM-DD')
and not EXISTS
(SELECT 1
FROM t_policy_change tpc
WHERE tpc.POLICY_ID = a.POLICY_ID
AND tpc.change_status = 3
and tpc.service_id = 82)
and a.organ_id in
(select organ_id
from t_company_organ
start with organ_id = '1'
connect by parent_id = prior organ_id)) table_a
where rownum < 21)
where my_rownum >= 1
SQL> select * from table(dbms_xplan.display);
PLAN_TABLE_OUTPUT
----------------------------------------------------------------------------------------------------------------------------
--------------------
-----------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)|
-----------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 475 | 81 (0)|
|* 1 | VIEW | | 1 | 475 | |
|* 2 | COUNT STOPKEY | | | | |
|* 3 | FILTER | | | | |
| 4 | NESTED LOOPS SEMI | | 1 | 268 | 81 (0)|
| 5 | NESTED LOOPS | | 1 | 258 | 80 (0)|
|* 6 | FILTER | | | | |
| 7 | NESTED LOOPS OUTER | | | | |
| 8 | NESTED LOOPS | | 1 | 178 | 78 (0)|
| 9 | NESTED LOOPS | | 1 | 114 | 77 (0)|
| 10 | VIEW | VW_NSO_1 | 7 | 154 | |
| 11 | SORT UNIQUE | | 7 | 70 | |
|* 12 | FILTER | | | | |
|* 13 | CONNECT BY WITH FILTERING | | | | |
| 14 | NESTED LOOPS | | | | |
|* 15 | INDEX UNIQUE SCAN | PK_T_COMPANY_ORGAN | 1 | 6 | |
| 16 | TABLE ACCESS BY USER ROWID| T_COMPANY_ORGAN | | | |
| 17 | NESTED LOOPS | | | | |
| 18 | BUFFER SORT | | 7 | 70 | |
| 19 | CONNECT BY PUMP | | | | |
|* 20 | INDEX RANGE SCAN | T_COMPANY_ORGAN_IDX_002 | 7 | 70 | 1 (0)|
|* 21 | TABLE ACCESS BY INDEX ROWID | T_CONTRACT_MASTER | 1 | 92 | 9 (0)|
|* 22 | INDEX RANGE SCAN | IDX_TCM_ORGAN_PID_SELLWAY | 1 | | 85 (0)|
|* 23 | INDEX RANGE SCAN | IDX_AGENCY_PROLICY_PROBLEM_POL | 1 | 6 | 1 (0)|
| 24 | SORT AGGREGATE | | 1 | 10 | |
|* 25 | TABLE ACCESS BY INDEX ROWID | T_POLICY_PROBLEM | 1 | 10 | 2 (50)|
|* 26 | INDEX RANGE SCAN | IDX_POLICY_PROBLEM__POLICY_ID | 1 | | 3 (0)|
| 27 | SORT AGGREGATE | | 1 | 10 | |
|* 28 | TABLE ACCESS BY INDEX ROWID | T_POLICY_PROBLEM | 1 | 10 | 2 (50)|
|* 29 | INDEX RANGE SCAN | IDX_POLICY_PROBLEM__POLICY_ID | 1 | | 3 (0)|
| 30 | SORT AGGREGATE | | 1 | 18 | |
| 31 | INLIST ITERATOR | | | | |
| 32 | TABLE ACCESS BY INDEX ROWID| T_IMAGE | 1 | 18 | 2 (50)|
|* 33 | INDEX RANGE SCAN | IDX_IMAGE__POLICY_TYPE | 1 | | 3 (0)|
|* 34 | TABLE ACCESS BY INDEX ROWID | T_POLICY_CHANGE | 1 | 11 | 2 (0)|
|* 35 | INDEX RANGE SCAN | IDX_POLICY_CHANGE__POLICY_ID | 12 | | 3 (0)|
| 36 | TABLE ACCESS BY INDEX ROWID | T_POLICY | 1 | 64 | 2 (50)|
|* 37 | INDEX UNIQUE SCAN | PK_T_POLICY | 1 | | 1 (0)|
| 38 | TABLE ACCESS BY INDEX ROWID | T_POLICY_REENTRY | 1 | 37 | 2 (50)|
|* 39 | INDEX RANGE SCAN | IDX_TPR_PID | 1 | | 1 (0)|
| 40 | TABLE ACCESS BY INDEX ROWID | T_COMPANY_ORGAN | 1 | 43 | 2 (50)|
|* 41 | INDEX UNIQUE SCAN | PK_T_COMPANY_ORGAN | 1 | | |
| 42 | INLIST ITERATOR | | | | |
|* 43 | INDEX RANGE SCAN | IDX_IMAGE__POLICY_TYPE | 833K| 8142K| 2 (0)|
| 44 | SORT AGGREGATE | | 1 | 9 | |
| 45 | TABLE ACCESS BY INDEX ROWID | T_POLICY_REENTRY | 1 | 9 | 2 (50)|
|* 46 | INDEX RANGE SCAN | IDX_TPR_PID | 1 | | 1 (0)|
-----------------------------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter("from_subquery_001"."MY_ROWNUM">=1)
2 - filter(ROWNUM<21)
3 - filter("PR"."REENTRY_ID" IS NULL OR "PR"."REENTRY_ID"= (SELECT /*+ */ MAX("TPR"."REENTRY_ID") FROM
"T_POLICY_REENTRY" "TPR" WHERE "TPR"."POLICY_ID"=:B1))
6 - filter("PR"."REENTRY_ID" IS NULL)
12 - filter(('2'='3' OR '3'='3' OR '12'='3' OR '18'='3') AND ('3'='2' OR '3'='3' OR '3'='12' OR '3'='18') AND
('2'='3' OR '3'='3' OR '12'='3' OR '18'='3'))
13 - filter("T_COMPANY_ORGAN"."ORGAN_ID"='1')
15 - access("T_COMPANY_ORGAN"."ORGAN_ID"='1')
20 - access("T_COMPANY_ORGAN"."PARENT_ID"=NULL)
21 - filter("SYS_ALIAS_13"."AGENCY_UW_STATE"<>'2')
22 - access("SYS_ALIAS_13"."ORGAN_ID"="VW_NSO_1"."$nso_col_1" AND "SYS_ALIAS_13"."SELL_WAY"='3')
filter("SYS_ALIAS_13"."SELL_WAY"='3' AND ("SYS_ALIAS_13"."SALE_CHANNEL"='3' OR
"SYS_ALIAS_13"."SALE_CHANNEL"='5') AND NOT EXISTS (SELECT /*+ */ 0 FROM "T_AGENCY_PROLICY_PROBLEM" "TAP" WHER
"TAP"."POLICY_ID"=:B1) AND (SELECT /*+ */ COUNT(*) FROM "T_POLICY_PROBLEM" "TPP" WHERE "TPP"."POLICY_ID"=:B2
AND TO_NUMBER("TPP"."ORIGIN_TYPE")=802)=0 AND (SELECT /*+ */ COUNT(*) FROM "T_POLICY_PROBLEM" "TPP" WHERE
"TPP"."POLICY_ID"=:B3 AND TO_NUMBER("TPP"."ORIGIN_TYPE")=801)=0 AND (SELECT /*+ */ TRUNC(MIN("TI"."SCAN_TIME"
FROM "T_IMAGE" "TI" WHERE "TI"."POLICY_ID"=:B4 AND ("TI"."IMAGE_TYPE_ID"=308 OR
"TI"."IMAGE_TYPE_ID"=309))>=TO_DATE('2011-01-01 00:00:00', 'yyyy-mm-dd hh24:mi:ss') AND NOT EXISTS (SELECT /*
*/ 0 FROM "T_POLICY_CHANGE" "TPC" WHERE "TPC"."POLICY_ID"=:B5 AND TO_NUMBER("TPC"."CHANGE_STATUS")=3 AND
"TPC"."SERVICE_ID"=82))
23 - access("TAP"."POLICY_ID"=:B1)
25 - filter(TO_NUMBER("TPP"."ORIGIN_TYPE")=802)
26 - access("TPP"."POLICY_ID"=:B1)
28 - filter(TO_NUMBER("TPP"."ORIGIN_TYPE")=801)
29 - access("TPP"."POLICY_ID"=:B1)
33 - access(("TI"."IMAGE_TYPE_ID"=308 OR "TI"."IMAGE_TYPE_ID"=309) AND "TI"."POLICY_ID"=:B1)
34 - filter(TO_NUMBER("TPC"."CHANGE_STATUS")=3 AND "TPC"."SERVICE_ID"=82)
35 - access("TPC"."POLICY_ID"=:B1)
37 - access("SYS_ALIAS_13"."POLICY_ID"="SYS_ALIAS_6"."POLICY_ID")
39 - access("PR"."POLICY_ID"(+)="SYS_ALIAS_13"."POLICY_ID")
41 - access("SYS_ALIAS_13"."ORGAN_ID"="E"."ORGAN_ID")
43 - access(("TII"."IMAGE_TYPE_ID"=308 OR "TII"."IMAGE_TYPE_ID"=309) AND
"TII"."POLICY_ID"="SYS_ALIAS_13"."POLICY_ID")
46 - access("TPR"."POLICY_ID"=:B1)
91 rows selected.
Statistics
----------------------------------------------------------
0 recursive calls
0 db block gets
2186915 consistent gets
12 physical reads
0 redo size
5202 bytes sent via SQL*Net to client
3835 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
8 sorts (memory)
0 sorts (disk)
13 rows processed
盖尔说，这个SQL要跑13秒，逻辑读200W左右，并且这个SQL上面有热点块，经常latch free ，呵呵他的DB是9i,latch free肯定是 cache buffers chains了
逻辑读太高了，有cache buffers chains 这个太正常不过了，很多时候所谓的热点块其实是SQL执行计划没走对产生的，这个SQL主要的表信息如下
SQL> select count(0) from t_contract_master;
COUNT(0)
----------
1131194
SQL> select count(0) from t_company_organ ;
COUNT(0)
----------
82
SQL> select count(0) from t_policy;
COUNT(0)
----------
1214045
SQL> select count(0) from t_policy_reentry;
COUNT(0)
----------
106262
索引信息我就不贴了，其实贴了也没啥用，反正我做SQL优化，单单看看执行计划就知道哪里有性能瓶颈，信不信由你呵呵。
这个SQL的性能瓶颈再22步，信不信由你，由于这个SQL有not exists 加上 and not exists 再加上 or .... ，没办法了，CBO 不能进行子查询展开，所以要优化这个SQL就得改写SQL。另外 还有这里 TO_NUMBER("TPC"."CHANGE_STATUS")=3 奶奶的， 写SQL的人 为啥 不写 where TPC.CHANGE_STATUS='3' 还让DB 自己转换一下，写那SQL的人真的很坑爹，不过你不要高兴，这个地方不是性能瓶颈，可以说这里对整体SQL并没什么影响，我只是提一下。由于无法连接他DB，也不知道业务逻辑，叫他把SQL用到的表给export出来也不行(被他老大发现了，说这个是机密
哈哈哈)，所以就不给他改SQL了，这个SQL要改正的地方有2个，1 是分页的地方，可以在里面分页，不过这个也无所谓，另外就是select count(*)
from t_policy_problem tpp
where tpp.origin_type = 801
and tpp.policy_id = a.policy_id) = 0
这个地方可以用exists改写，不过我估计效率也提升不了多少，这个SQL最大的性能瓶颈还是在于22步，怎么样才能把第22步给干掉呢？第22步骤其实是和第10步骤做嵌套循环，oh, 我的妈啊，本来第22步骤性能都这么差了，你还来给我嵌套循环搞几次，那步搞死人，逻辑读不大才怪呢，所以 我让盖尔 查看
select organ_id
from t_company_organ
start with organ_id = '1'
connect by parent_id = prior organ_id
返回多少行，盖尔说上面的SQL返回82行，恩，心里有底了，这个子查询不应该被展开，它才返回82条记录，应该直接让它走filter,所以加了个HINT ，代码和执行计划如下
select *
from (select rownum as my_rownum, table_a.*
from (select e.company_name organ_name,
e.abbr_name organ_abbr_name,
a.agency_uw_state,
a.agency_uw_time,
a.ht_bill_def_version,
(SELECT version.ht_bill_def_version_type_id
FROM t_ht_bill_def_version version
WHERE a.ht_bill_def_version =
version.ht_bill_def_version) ht_bill_def_version_type_id,
a.policy_id,
tp.p_state_id,
a.organ_id,
a.agent_id,
tp.send_code,
a.policy_code,
a.insert_time,
(SELECT tct.real_name
FROM t_customer tct
WHERE tct.customer_id = a.applicant_id) holder_name,
(SELECT tbe.bank_code
FROM t_bank_employee tbe
WHERE a.agency_hand = tbe.emp_id) bank_code,
a.agency_code,
(SELECT b.is_charge
FROM t_agent b
WHERE a.agent_id = b.agent_id) is_charge,
(SELECT b.real_name
FROM t_agent b
WHERE a.agent_id = b.agent_id) agent_name,
tp.record_end,
tp.inspect_time,
tp.record_insert_time,
a.derivation,
tp.pending_cause,
a.update_time,
tp.scan_time,
tp.first_end,
tp.priority_id,
a.policy_type,
(SELECT f.real_name
FROM t_employee f
WHERE tp.first_user_id = f.emp_id) first_name,
'' recorder_name,
(SELECT g.real_name
FROM t_employee g
WHERE a.agency_uw_rec_id = g.emp_id) checker_name,
pr.apply_time,
pr.finish_time,
pr.handler_id,
pr.apply_id
from t_contract_master a,
t_company_organ e,
t_policy tp,
t_policy_reentry pr
where a.organ_id = e.organ_id
and a.policy_id = tp.policy_id
and pr.policy_id(+) = a.policy_id
and (pr.reentry_id =
(select max(tpr.reentry_id)
from t_policy_reentry tpr
where tpr.policy_id = a.policy_id) or
pr.reentry_id is null)
and pr.reentry_id is null
and a.sell_way in ('2', '3', '12', '18')
and a.agency_uw_state <> '2'
and not EXISTS (SELECT 1
FROM t_agency_prolicy_problem tap
WHERE tap.POLICY_ID = a.POLICY_ID)
and (select count(*)
from t_policy_problem tpp
where tpp.origin_type = 802
and tpp.policy_id = a.policy_id) = 0
and (select count(*)
from t_policy_problem tpp
where tpp.origin_type = 801
and tpp.policy_id = a.policy_id) = 0
and a.sell_way = '3'
and a.sale_channel in ('3', '5')
and EXISTS (SELECT 1
FROM t_image tii
WHERE tii.POLICY_ID = a.POLICY_ID
and tii.image_type_id in (308, 309))
and (select trunc(min(ti.scan_time))
from t_image ti
WHERE ti.image_type_id in (308, 309)
and ti.policy_id = a.policy_id) >=
to_date('2011-01-01', 'YYYY-MM-DD')
and not EXISTS
(SELECT 1
FROM t_policy_change tpc
WHERE tpc.POLICY_ID = a.POLICY_ID
AND tpc.change_status = 3
and tpc.service_id = 82)
and a.organ_id in
(select /*+ NO_UNNEST */ organ_id
from t_company_organ
start with organ_id = '1'
connect by parent_id = prior organ_id)) table_a
where rownum < 21)
where my_rownum >= 1
Statistics
----------------------------------------------------------
0 recursive calls
0 db block gets
261250 consistent gets
0 physical reads
0 redo size
5202 bytes sent via SQL*Net to client
3847 bytes received via SQL*Net from client
2 SQL*Net roundtrips to/from client
364 sorts (memory)
0 sorts (disk)
13 rows processed
SQL> select * from table(dbms_xplan.display);
PLAN_TABLE_OUTPUT
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
| Id | Operation | Name | Rows | Bytes | Cost (%CPU)|
----------------------------------------------------------------------------------------------------------------
| 0 | SELECT STATEMENT | | 1 | 475 | 277 (2)|
|* 1 | VIEW | | 1 | 475 | |
|* 2 | COUNT STOPKEY | | | | |
|* 3 | FILTER | | | | |
| 4 | NESTED LOOPS SEMI | | 1 | 248 | 267 (0)|
| 5 | NESTED LOOPS | | 1 | 238 | 266 (0)|
|* 6 | FILTER | | | | |
| 7 | NESTED LOOPS OUTER | | | | |
| 8 | NESTED LOOPS | | 1 | 135 | 264 (0)|
|* 9 | TABLE ACCESS BY INDEX ROWID | T_CONTRACT_MASTER | 1 | 92 | 263 (0)|
|* 10 | INDEX SKIP SCAN | IDX_CONT_MA__ORGAN_SELL | 3144 | | 782 (0)|
|* 11 | FILTER | | | | |
|* 12 | CONNECT BY WITH FILTERING | | | | |
| 13 | NESTED LOOPS | | | | |
|* 14 | INDEX UNIQUE SCAN | PK_T_COMPANY_ORGAN | 1 | 6 | |
| 15 | TABLE ACCESS BY USER ROWID| T_COMPANY_ORGAN | | | |
| 16 | NESTED LOOPS | | | | |
| 17 | BUFFER SORT | | 7 | 70 | |
| 18 | CONNECT BY PUMP | | | | |
|* 19 | INDEX RANGE SCAN | T_COMPANY_ORGAN_IDX_002 | 7 | 70 | 1 (0)|
| 20 | TABLE ACCESS BY INDEX ROWID | T_COMPANY_ORGAN | 1 | 43 | 2 (50)|
|* 21 | INDEX UNIQUE SCAN | PK_T_COMPANY_ORGAN | 1 | | |
| 22 | TABLE ACCESS BY INDEX ROWID | T_POLICY_REENTRY | 1 | 39 | 2 (50)|
|* 23 | INDEX RANGE SCAN | IDX_TPR_PID | 1 | | 1 (0)|
| 24 | TABLE ACCESS BY INDEX ROWID | T_POLICY | 1 | 64 | 2 (50)|
|* 25 | INDEX UNIQUE SCAN | PK_T_POLICY | 1 | | 1 (0)|
| 26 | INLIST ITERATOR | | | | |
|* 27 | INDEX RANGE SCAN | IDX_IMAGE__POLICY_TYPE | 356K| 3477K| 2 (0)|
| 28 | SORT AGGREGATE | | 1 | 11 | |
| 29 | TABLE ACCESS BY INDEX ROWID | T_POLICY_REENTRY | 1 | 11 | 2 (50)|
|* 30 | INDEX RANGE SCAN | IDX_TPR_PID | 1 | | 1 (0)|
|* 31 | INDEX RANGE SCAN | IDX_AGENCY_PROLICY_PROBLEM_POL | 1 | 6 | 1 (0)|
| 32 | SORT AGGREGATE | | 1 | 10 | |
|* 33 | TABLE ACCESS BY INDEX ROWID | T_POLICY_PROBLEM | 1 | 10 | 2 (50)|
|* 34 | INDEX RANGE SCAN | IDX_POLICY_PROBLEM__POLICY_ID | 1 | | 3 (0)|
| 35 | SORT AGGREGATE | | 1 | 10 | |
|* 36 | TABLE ACCESS BY INDEX ROWID | T_POLICY_PROBLEM | 1 | 10 | 2 (50)|
|* 37 | INDEX RANGE SCAN | IDX_POLICY_PROBLEM__POLICY_ID | 1 | | 3 (0)|
| 38 | SORT AGGREGATE | | 1 | 18 | |
| 39 | INLIST ITERATOR | | | | |
| 40 | TABLE ACCESS BY INDEX ROWID | T_IMAGE | 1 | 18 | 2 (50)|
|* 41 | INDEX RANGE SCAN | IDX_IMAGE__POLICY_TYPE | 1 | | 3 (0)|
|* 42 | TABLE ACCESS BY INDEX ROWID | T_POLICY_CHANGE | 1 | 13 | 2 (0)|
|* 43 | INDEX RANGE SCAN | IDX_POLICY_CHANGE__POLICY_ID | 12 | | 3 (0)|
----------------------------------------------------------------------------------------------------------------
Predicate Information (identified by operation id):
---------------------------------------------------
1 - filter("from_subquery_001"."MY_ROWNUM">=1)
2 - filter(ROWNUM<21)
3 - filter(("PR"."REENTRY_ID" IS NULL OR "PR"."REENTRY_ID"= (SELECT /*+ */ MAX("TPR"."REENTRY_ID") FROM
"T_POLICY_REENTRY" "TPR" WHERE "TPR"."POLICY_ID"=:B1)) AND NOT EXISTS (SELECT /*+ */ 0 FROM
"T_AGENCY_PROLICY_PROBLEM" "TAP" WHERE "TAP"."POLICY_ID"=:B2) AND (SELECT /*+ */ COUNT(*) FROM
"T_POLICY_PROBLEM" "TPP" WHERE "TPP"."POLICY_ID"=:B3 AND TO_NUMBER("TPP"."ORIGIN_TYPE")=802)=0 AND (SELECT /*+
*/ COUNT(*) FROM "T_POLICY_PROBLEM" "TPP" WHERE "TPP"."POLICY_ID"=:B4 AND TO_NUMBER("TPP"."ORIGIN_TYPE")=801)=0
AND (SELECT /*+ */ TRUNC(MIN("TI"."SCAN_TIME")) FROM "T_IMAGE" "TI" WHERE "TI"."POLICY_ID"=:B5 AND
("TI"."IMAGE_TYPE_ID"=308 OR "TI"."IMAGE_TYPE_ID"=309))>=TO_DATE('2011-01-01 00:00:00', 'yyyy-mm-dd
hh24:mi:ss') AND NOT EXISTS (SELECT /*+ */ 0 FROM "T_POLICY_CHANGE" "TPC" WHERE "TPC"."POLICY_ID"=:B6 AND
TO_NUMBER("TPC"."CHANGE_STATUS")=3 AND "TPC"."SERVICE_ID"=82))
6 - filter("PR"."REENTRY_ID" IS NULL)
9 - filter("SYS_ALIAS_14"."SALE_CHANNEL"='3' OR "SYS_ALIAS_14"."SALE_CHANNEL"='5')
10 - access("SYS_ALIAS_14"."SELL_WAY"='3')
filter("SYS_ALIAS_14"."SELL_WAY"='3' AND "SYS_ALIAS_14"."AGENCY_UW_STATE"<>'2' AND EXISTS (SELECT /*+
NO_UNNEST */ 0 FROM "T_COMPANY_ORGAN" "T_COMPANY_ORGAN" WHERE "T_COMPANY_ORGAN"."PARENT_ID"=NULL AND
("T_COMPANY_ORGAN"."ORGAN_ID"=:B1)))
11 - filter("T_COMPANY_ORGAN"."ORGAN_ID"=:B1)
12 - filter("T_COMPANY_ORGAN"."ORGAN_ID"='1')
14 - access("T_COMPANY_ORGAN"."ORGAN_ID"='1')
19 - access("T_COMPANY_ORGAN"."PARENT_ID"=NULL)
21 - access("SYS_ALIAS_14"."ORGAN_ID"="E"."ORGAN_ID")
23 - access("PR"."POLICY_ID"(+)="SYS_ALIAS_14"."POLICY_ID")
25 - access("SYS_ALIAS_14"."POLICY_ID"="SYS_ALIAS_6"."POLICY_ID")
27 - access(("TII"."IMAGE_TYPE_ID"=308 OR "TII"."IMAGE_TYPE_ID"=309) AND
"TII"."POLICY_ID"="SYS_ALIAS_14"."POLICY_ID")
30 - access("TPR"."POLICY_ID"=:B1)
31 - access("TAP"."POLICY_ID"=:B1)
33 - filter(TO_NUMBER("TPP"."ORIGIN_TYPE")=802)
34 - access("TPP"."POLICY_ID"=:B1)
36 - filter(TO_NUMBER("TPP"."ORIGIN_TYPE")=801)
37 - access("TPP"."POLICY_ID"=:B1)
41 - access(("TI"."IMAGE_TYPE_ID"=308 OR "TI"."IMAGE_TYPE_ID"=309) AND "TI"."POLICY_ID"=:B1)
42 - filter(TO_NUMBER("TPC"."CHANGE_STATUS")=3 AND "TPC"."SERVICE_ID"=82)
43 - access("TPC"."POLICY_ID"=:B1)
88 rows selected.
这样之后逻辑读整整下降了进10倍，SQL也能1秒钟跑完，不过逻辑读还是有点大，依然有20W左右，如果这个SQL执行频率很高，那么系统肯定又会出现所谓的热点块了
这个时候SQL的性能瓶颈在第3步，信不信由你，要再继续优化这个SQL ，我就需要连接他的DB了，不过无法连接，很遗憾，另外这个SQL的索引也可以稍微调节一下，我估计都弄好之后，这个SQL的逻辑读应该降低到1万到几万之间，本来还想继续搞这个SQL的，但是盖尔说搞另外一个，奶奶的又扔给我一个变态SQL，好吧，这个SQL的优化就到此结束。
欢迎广大网友给出继续优化这个SQL的思路 ，看你想的是否和我一样(方法多多哦)，大家尽情发挥。