sql 查询需要很多时间

Question

I am running the following query:

SELECT * 
  FROM dblappc.credit_history 
 WHERE crd_effective_date > TO_DATE('20100801','YYYYMMDD') 
   AND o_crd_score > 650 
   AND crd_expiration_date IS NULL

• Above table has no indexes
• crd_expiration_date is part of the primary key

How can I speed up the query?
Could I use parallel hint here so that to give 500 rows at a time at least?

Below is the table structure:

CREATE TABLE DFQAPP13.CREDIT_HISTORY
(
  BAN                  NUMBER(9) CONSTRAINT CRHST_BAN_NN NOT NULL,
  CRD_SEQ_NO           NUMBER(9) CONSTRAINT CRDHST_CRD_SEQ_NO_NN NOT NULL,
  SYS_CREATION_DATE    DATE                     NOT NULL,
  SYS_UPDATE_DATE      DATE,
  OPERATOR_ID          NUMBER(9),
  APPLICATION_ID       CHAR(6 BYTE),
  DL_SERVICE_CODE      CHAR(5 BYTE),
  DL_UPDATE_STAMP      NUMBER(4),
  CRD_EFFECTIVE_DATE   DATE CONSTRAINT CRDHST_CRD_EFFECTIVE_DATE_NN NOT NULL,
  CRD_EXPIRATION_DATE  DATE,
  CRD_VET_TYPE         CHAR(4 BYTE),
  O_CRD_APPLIC_NUM     NUMBER(9),
  O_CRD_DECISION       CHAR(2 BYTE),
  O_CRD_SCORE          NUMBER(7),
  O_CRD_POLICY_RULE1   VARCHAR2(40 BYTE),
  O_CRD_POLICY_RULE2   VARCHAR2(40 BYTE),
  O_CRD_POLICY_RULE3   VARCHAR2(40 BYTE),
  O_CRD_POLICY_RULE4   VARCHAR2(40 BYTE),
  O_CRD_POLICY_RULE5   VARCHAR2(40 BYTE),
  O_CRD_POLICY_RULE6   VARCHAR2(40 BYTE),
  CRD_CLASS            CHAR(1 BYTE),
  CRD_CLASS_CHG_TYPE   CHAR(1 BYTE),
  CRD_CHG_RSN_TEXT     CHAR(100 BYTE),
  I_CRD_REQ_CTN_QTY    NUMBER(7),
  CRD_APR_CTN_QTY      NUMBER(7),
  I_CRD_BANK_BRANCH    VARCHAR2(100 BYTE),
  I_CRD_TACT_BANK_CD   CHAR(1 BYTE),
  I_CRD_BANK_DATE      DATE,
  I_ESAT_CUST_IND      CHAR(1 BYTE),
  O_DUNS_RET_CODE1     CHAR(4 BYTE),
  O_DUNS_RET_CODE2     CHAR(4 BYTE),
  O_DUNS_RET_NUM       VARCHAR2(18 BYTE),
  O_DUNS_NUM           NUMBER(9),
  O_DUNS_FIN_STRENGTH  CHAR(3 BYTE),
  O_DUNS_COMP_COND     CHAR(1 BYTE),
  O_DUNS_PAYM_SCORE    NUMBER(4),
  O_DUNS_CCJ1_EIRE     NUMBER(6),
  O_DUNS_CCJ2_EIRE     NUMBER(6),
  O_DUNS_CCJ3_EIRE     NUMBER(6),
  O_DUNS_CCJ4_EIRE     NUMBER(6),
  O_DUNS_CCJ5_EIRE     NUMBER(6),
  O_DUNS_CCJ1_UK       NUMBER(4),
  O_DUNS_CCJ2_UK       NUMBER(4),
  O_DUNS_CCJ3_UK       NUMBER(4),
  O_DUNS_CCJ4_UK       NUMBER(4),
  O_DUNS_CCJ5_UK       NUMBER(4),
  I_PHONE_TYPE         CHAR(3 BYTE),
  I_PAID_ENHANCE_NUM   NUMBER(1),
  I_CHURN_CUST_IND     CHAR(1 BYTE),
  I_EX_DIRECTORY_IND   CHAR(1 BYTE),
  I_ITEMISED_BIL_IND   CHAR(1 BYTE),
  CONV_RUN_NO          NUMBER(3)
)

Answer 1

Without knowing the size of the table, the existing indexes, the execution plan and other details, it's hard to give any advice.

Since the query is searching in two ranges, it's not easy to get all results using only indexes.

But I would try first these 2 options:

• Simple index on crd_effective_date and compound on (crd_expiration_date, o_crd_score) or

• Simple index on o_crd_score and compound on (crd_expiration_date, crd_effective_date)

You can check the execution plan of a query (now, without indexes and after adding them), using:EXPLAIN PLAN

Answer 2

There are one specific issue and some general improvements.

Firstly, null values not indexed, therefore Oracle most probably choose to full index scan or index skip scan variants for your query with index build above primary key. If you really need to include crd_expiration_date into primary key, create another index for this query on o_crd_score and crd_effective_date fields only.

create index X_CREDIT_HISTORY_DATE_SCORE 
  on CREDIT_HISTORY (o_crd_score, crd_effective_date)

If after that Oracle won't use new index force it in query text:

select /*+ index(hist_data X_CREDIT_HISTORY_DATE_SCORE) */ 
  * 
from 
  dblappc.credit_history hist_data
where 
  crd_effective_date>to_date('20100801','YYYYMMDD') 
  and 
  o_crd_score >650 
  and 
  crd_expiration_date is null

General issues are common and mentioned in previous answers:

1. Select only specific fields, which you really need;
2. Avoid using nulls in primary keys;
3. Surrogate primary keys are preferred.

Update

Oooops... I missed phrase "Above table has no indexes" while reading question before.

Therefore only one recommendation: just create an index.

Answer 3

You can make it faster by

• SELECT the specific item you want to reduce the result weight.
• In WHERE clause use the o_crd_score >650 in first place, crd_expiration_date is null in second place and crd_effective_date>to_date('20100801','YYYYMMDD') in the last

So the query will be like.

SELECT a,b,d... 
FROM dblappc.credit_history
WHERE 
  o_crd_score >650 and 
  crd_expiration_date is null
  crd_effective_date>to_date('20100801','YYYYMMDD') and 

What I am doing here is that short circuiting the the records to the minimum so that the date check will be applied to the filtered out record.

Also you can make it fast by applying the INDEX on the table.