Query is taking long time to execute after migrating to 10g r2

Question

Hi
We recently migrated the database from 9i to 10gr2 ((10.2.0.2.0).. This query was running in acceptable time before the upgrade in 9i.. Now it is taking a long long time to execute this... Can you please let me know what should i do to improve the performance now.. We are running stats everyday..
Thanks for your help,
Shree
======================================================================================
SELECT cr.cash_receipt_id
,cr.pay_from_customer
,cr.receipt_number
,cr.receipt_date
,cr.amount
,cust.account_number
,crh.gl_date
,cr.set_of_books_id
,sum(ra.amount_applied) amount_applied
FROM AR_CASH_RECEIPTS_ALL cr
,AR_RECEIVABLE_APPLICATIONS_ALL ra
,hz_cust_accounts cust
,AR_CASH_RECEIPT_HISTORY_ALL crh
,GL_PERIOD_STATUSES gps
,FND_APPLICATION app
WHERE cr.cash_receipt_id = ra.cash_receipt_id
AND ra.status = 'UNAPP'
AND cr.status <> 'REV'
AND cust.cust_account_id = cr.pay_from_customer
AND substr(cust.account_number,1,2) <> 'SI' -- Don't allocate Unapplied receipts FOR SI customers
AND crh.cash_receipt_id = cr.cash_receipt_id
AND app.application_id = gps.application_id
AND app.application_short_name = 'AR'
AND gps.period_name = 'May-07'
AND crh.gl_date <= gps.end_date
AND cr.receipt_number not like 'WH%'
-- AND cust.customer_number = '0000079260001'
GROUP BY cr.cash_receipt_id
,cr.pay_from_customer
,cr.receipt_number
,cr.receipt_date
,cr.amount
,cust.account_number
,crh.gl_date
,cr.set_of_books_id
HAVING sum(ra.amount_applied) > 0;
=========================================================================================
Here is the explain plan in 10g r2 (10.2.0.2.0)
PLAN_TABLE_OUTPUT
Plan hash value: 2617075047
| Id | Operation | Name | Rows | Bytes |TempSpc| Cost (%CPU)|
| 0 | SELECT STATEMENT | | 92340 | 10M| | 513K (1)|
|* 1 | FILTER | | | | | |
| 2 | HASH GROUP BY | | 92340 | 10M| 35M| 513K (1)|
| 3 | TABLE ACCESS BY INDEX ROWID | AR_RECEIVABLE_APPLICATIONS_ALL | 2 | 34 |
| 4 | NESTED LOOPS | | 184K| 21M| | 510K (1)|
|* 5 | HASH JOIN | | 99281 | 9M| 3296K| 176K (1)|
|* 6 | TABLE ACCESS FULL | HZ_CUST_ACCOUNTS | 112K| 1976K| | 22563 (1)|
|* 7 | HASH JOIN | | 412K| 33M| 25M| 151K (1)|
| 8 | TABLE ACCESS BY INDEX ROWID | AR_CASH_RECEIPT_HISTORY_ALL | 332K| 4546K|
| 9 | NESTED LOOPS | | 498K| 19M| | 26891 (1)|
| 10 | NESTED LOOPS | | 2 | 54 | | 4 (0)|
| 11 | TABLE ACCESS BY INDEX ROWID| FND_APPLICATION | 1 | 8 | | 1 (0)|
|* 12 | INDEX UNIQUE SCAN | FND_APPLICATION_U3 | 1 | | | 0 (0)|
| 13 | TABLE ACCESS BY INDEX ROWID| GL_PERIOD_STATUSES | 2 | 38 | | 3 (0)
|* 14 | INDEX RANGE SCAN | GL_PERIOD_STATUSES_U1 | 1 | | | 2 (0)|
|* 15 | INDEX RANGE SCAN | AR_CASH_RECEIPT_HISTORY_N2 | 332K| | | 1011 (1)
PLAN_TABLE_OUTPUT
|* 16 | TABLE ACCESS FULL | AR_CASH_RECEIPTS_ALL | 5492K| 235M| | 108K
|* 17 | INDEX RANGE SCAN | AR_RECEIVABLE_APPLICATIONS_N1 | 4 | | | 2
Predicate Information (identified by operation id):
1 - filter(SUM("RA"."AMOUNT_APPLIED")>0)
5 - access("CUST"."CUST_ACCOUNT_ID"="CR"."PAY_FROM_CUSTOMER")
6 - filter(SUBSTR("CUST"."ACCOUNT_NUMBER",1,2)<>'SI')
7 - access("CRH"."CASH_RECEIPT_ID"="CR"."CASH_RECEIPT_ID")
12 - access("APP"."APPLICATION_SHORT_NAME"='AR')
14 - access("APP"."APPLICATION_ID"="GPS"."APPLICATION_ID" AND "GPS"."PERIOD_NAME"='May-07')
filter("GPS"."PERIOD_NAME"='May-07')
15 - access("CRH"."GL_DATE"<="GPS"."END_DATE")
16 - filter("CR"."STATUS"<>'REV' AND "CR"."RECEIPT_NUMBER" NOT LIKE 'WH%')
17 - access("CR"."CASH_RECEIPT_ID"="RA"."CASH_RECEIPT_ID" AND "RA"."STATUS"='UNAPP')
filter("RA"."CASH_RECEIPT_ID" IS NOT NULL)
Here is the explain plan in 9i
Execution Plan
0 SELECT STATEMENT Optimizer=CHOOSE (Cost=445977 Card=78530 By
tes=9423600)
1 0 FILTER
2 1 SORT (GROUP BY) (Cost=445977 Card=78530 Bytes=9423600)
3 2 HASH JOIN (Cost=443717 Card=157060 Bytes=18847200)
4 3 HASH JOIN (Cost=99563 Card=94747 Bytes=9758941)
5 4 TABLE ACCESS (FULL) OF 'HZ_CUST_ACCOUNTS' (Cost=12
286 Card=110061 Bytes=1981098)
6 4 HASH JOIN (Cost=86232 Card=674761 Bytes=57354685)
7 6 TABLE ACCESS (BY INDEX ROWID) OF 'AR_CASH_RECEIP
T_HISTORY_ALL' (Cost=17532 Card=542304 Bytes=7592256)
8 7 NESTED LOOPS (Cost=17536 Card=809791 Bytes=332
01431)
9 8 NESTED LOOPS (Cost=4 Card=1 Bytes=27)
10 9 TABLE ACCESS (BY INDEX ROWID) OF 'FND_APPL
ICATION' (Cost=1 Card=1 Bytes=8)
11 10 INDEX (UNIQUE SCAN) OF 'FND_APPLICATION_
U3' (UNIQUE)
12 9 TABLE ACCESS (BY INDEX ROWID) OF 'GL_PERIO
D_STATUSES' (Cost=3 Card=1 Bytes=19)
13 12 INDEX (RANGE SCAN) OF 'GL_PERIOD_STATUSE
S_U1' (UNIQUE) (Cost=2 Card=1)
14 8 INDEX (RANGE SCAN) OF 'AR_CASH_RECEIPT_HISTO
RY_N2' (NON-UNIQUE) (Cost=1740 Card=542304)
15 6 TABLE ACCESS (FULL) OF 'AR_CASH_RECEIPTS_ALL' (C
ost=60412 Card=8969141 Bytes=394642204)
16 3 TABLE ACCESS (FULL) OF 'AR_RECEIVABLE_APPLICATIONS_A
LL' (Cost=337109 Card=15613237 Bytes=265425029)

Answer 1

Hi,
The plan between 9i and 10g is pretty the same but the amount of data fetched has considerably increased. I guess the query was performing slow even in 9i.
The AR_CASH_RECEIPT_HISTORY_ALL is presently having 332000 rows in 10g where as it was 17532 in 9i.
AR_CASH_RECEIPT_HISTORY_N2 is now having 332,000 rows in 10g where as in 9i it had 1,740
Try creating some indexes on
AR_CASH_RECEIPTS_ALL
hz_cust_accounts