如何在以下更新 sql 上進行 SQL 調優
[英]how to do SQL tuning on below update sql
問題描述
大家，最近我參與了一個oracle的SQL調優任務，我遇到了一個我認為非常困難的問題，我什至可以說我被這個問題嚇到了，我從DBA那里得到了AWR報告，並且似乎是來自AMR的紅線SQL需要做一些調整（我已經將這些SQL粘貼如下，這個sql在SP中）。但我不知道是什么原因導致性能不佳，任何人都可以幫助提供一些解決方案或調整SQL的想法？
如果您需要 AWR 的更多證據，請告訴我。
提前致謝...
UPDATE tax_ratio tar
SET
( ecm,
esm,
epm,
ecam,
update_dt,
update_by ) = (
SELECT
nvl(src_t1.ecm,0) AS ecm,
nvl(src_t1.esm,0) AS esm,
nvl(src_t1.epm,0) AS epm,
nvl(src_t1.ecam,0) AS ecam,
SYSDATE,
'ffee_user'
FROM
(
SELECT
city_code,
tax_type,
company_type,
taxpayer,
company_group,
company_tax_type,
SUM(new_tax_current_mth) /12 AS ecm,
SUM(new_tax_miss_current_mth) /12 AS esm,
SUM(new_tax_get_current_mth) /12 AS epm,
SUM(new_tax_special_current_mth) /12 AS ecam
FROM
tax_ratio
WHERE
city_code ='001'
AND company_type ='typ_01'
AND tax_mth <= add_months(TO_DATE('08-JUL-2015'),-3)
AND tax_mth >= add_months(TO_DATE('08-JUL-2015'),-14)
AND eff_date =TO_DATE('08-JUL-2015')
AND tax_type = '00'
GROUP BY
city_code,
tax_type,
company_type,
taxpayer,
company_group,
company_tax_type
HAVING SUM(new_tax_current_mth) <> 0
OR SUM(new_tax_miss_current_mth) <> 0
OR SUM(new_tax_get_current_mth) <> 0
OR SUM(new_tax_special_current_mth) <> 0
) src_t1
WHERE
tar.city_code = src_t1.city_code
AND tar.tax_type = src_t1.tax_type
AND tar.company_type = src_t1.company_type
AND tar.taxpayer = src_t1.taxpayer
AND nvl(tar.company_group,'-99999') = nvl(src_t1.company_group,'-99999')
AND (
src_t1.ecm IS NOT NULL
OR src_t1.esm IS NOT NULL
OR src_t1.epm IS NOT NULL
OR src_t1.ecam IS NOT NULL
)
AND tar.tax_mth =TO_DATE('08-JUL-2015')
AND tar.company_tax_type = src_t1.company_tax_type
)
WHERE
tar.city_code ='001'
AND tar.company_type ='typ_01'
AND tar.tax_mth =TO_DATE('08-JUL-2015')
AND EXISTS (
SELECT
1
FROM
(
SELECT
city_code,
tax_type,
company_type,
taxpayer,
company_group,
company_tax_type,
SUM(new_tax_current_mth) /12 AS ecm,
SUM(new_tax_miss_current_mth) /12 AS esm,
SUM(new_tax_get_current_mth) /12 AS epm,
SUM(new_tax_special_current_mth) /12 AS ecam
FROM
tax_ratio
WHERE
city_code ='001'
AND company_type ='typ_01'
AND tax_mth <= add_months(TO_DATE('08-JUL-2015'),-3)
AND tax_mth >= add_months(TO_DATE('08-JUL-2015'),-14)
AND eff_date =TO_DATE('08-Aug-2015')
AND tax_type = '00'
GROUP BY
city_code,
tax_type,
company_type,
taxpayer,
company_group,
company_tax_type
HAVING SUM(new_tax_current_mth) <> 0
OR SUM(new_tax_miss_current_mth) <> 0
OR SUM(new_tax_get_current_mth) <> 0
OR SUM(new_tax_special_current_mth) <> 0
) src_t1
WHERE
tar.city_code = src_t1.city_code
AND tar.tax_type = src_t1.tax_type
AND tar.company_type = src_t1.company_type
AND tar.taxpayer = src_t1.taxpayer
AND nvl(tar.company_group,'-99999') = nvl(src_t1.company_group,'-99999')
AND (
src_t1.ecm IS NOT NULL
OR src_t1.esm IS NOT NULL
OR src_t1.epm IS NOT NULL
OR src_t1.ecam IS NOT NULL
)
AND tar.tax_mth =TO_DATE('08-JUL-2015')
AND tar.company_tax_type = src_t1.company_tax_type
)
add the EXPLAIN PLAN
PLAN HASH VALUE: 3650439649
----------------------------------------------------------------------------------------------------------------------
| ID | OPERATION | NAME | ROWS | BYTES |TEMPSPC| COST (%CPU)| TIME |
----------------------------------------------------------------------------------------------------------------------
| 0 | UPDATE STATEMENT | | 1698 | 179K| | 6169K (1)| 00:08:02 |
| 1 | UPDATE | TAX_RATIO | | | | | |
|* 2 | HASH JOIN RIGHT SEMI | | 1698 | 179K| | 732K (1)| 00:00:58 |
| 3 | VIEW | | 39251 | 1111K| | 371K (2)| 00:00:29 |
|* 4 | FILTER | | | | | | |
| 5 | SORT GROUP BY | | 39251 | 2414K| 100M| 371K (2)| 00:00:29 |
|* 6 | TABLE ACCESS FULL | TAX_RATIO | 1140K| 68M| | 365K (2)| 00:00:29 |
|* 7 | TABLE ACCESS FULL | TAX_RATIO | 207K| 15M| | 361K (1)| 00:00:29 |
| 8 | VIEW | | 1 | 81 | | 484 (1)| 00:00:01 |
|* 9 | FILTER | | | | | | |
| 10 | SORT GROUP BY | | 1 | 63 | | 484 (1)| 00:00:01 |
|* 11 | FILTER | | | | | | |
|* 12 | TABLE ACCESS BY INDEX ROWID BATCHED| TAX_RATIO | 1 | 63 | | 483 (0)| 00:00:01 |
|* 13 | INDEX RANGE SCAN | TAX_RATIO_TAXPAYER_IDX | 544 | | | 3 (0)| 00:00:01 |
----------------------------------------------------------------------------------------------------------------------
PREDICATE INFORMATION (IDENTIFIED BY OPERATION ID):
---------------------------------------------------
2 - ACCESS("TAR"."CITY_CODE"="SRC_T1"."CITY_CODE" AND "TAR"."TAX_TYPE"="SRC_T1"."TAX_TYPE" AND
"TAR"."COMPANY_TYPE"="SRC_T1"."COMPANY_TYPE" AND "TAR"."TAXPAYER"="SRC_T1"."TAXPAYER" AND
NVL("TAR"."COMPANY_GROUP",'-99999')=NVL("SRC_T1"."COMPANY_GROUP",'-99999') AND "TAR"."COMPANY_TAX_TYPE"="SRC_T1"."COMPANY_TAX_TYPE")
4 - FILTER((SUM("NEW_TAX_CURRENT_MTH")<>0 OR SUM("NEW_TAX_MISS_CURRENT_MTH")<>0 OR SUM("NEW_TAX_GET_CURRENT_MTH")<>0 OR SUM("NEW_TAX_SPECIAL_CURRENT_MTH")<>0) AND
(SUM("NEW_TAX_CURRENT_MTH")/12 IS NOT NULL OR SUM("NEW_TAX_MISS_CURRENT_MTH")/12 IS NOT NULL OR SUM("NEW_TAX_GET_CURRENT_MTH")/12 IS NOT NULL OR
SUM("NEW_TAX_SPECIAL_CURRENT_MTH")/12 IS NOT NULL))
6 - FILTER("COMPANY_TYPE"='LIMIT' AND "TAX_TYPE"='00' AND "TAX_MTH">=TO_DATE(' 2017-03-31 00:00:00',
'SYYYY-MM-DD HH24:MI:SS') AND "TAX_MTH"<=TO_DATE(' 2018-02-28 00:00:00', 'SYYYY-MM-DD HH24:MI:SS') AND
"CITY_CODE"='001' AND "NEW_TAX_MISS_CURRENT_MTH"=TO_DATE(' 2200-01-01 00:00:00', 'SYYYY-MM-DD HH24:MI:SS'))
7 - FILTER("TAR"."TAX_MTH"=TO_DATE(' 2018-05-31 00:00:00', 'SYYYY-MM-DD HH24:MI:SS') AND
"TAR"."COMPANY_TYPE"='LIMIT' AND "TAR"."CITY_CODE"='001')
9 - FILTER((SUM("NEW_TAX_CURRENT_MTH")<>0 OR SUM("NEW_TAX_MISS_CURRENT_MTH")<>0 OR SUM("NEW_TAX_GET_CURRENT_MTH")<>0 OR SUM("NEW_TAX_SPECIAL_CURRENT_MTH")<>0) AND
(SUM("NEW_TAX_CURRENT_MTH")/12 IS NOT NULL OR SUM("NEW_TAX_MISS_CURRENT_MTH")/12 IS NOT NULL OR SUM("NEW_TAX_GET_CURRENT_MTH")/12 IS NOT NULL OR
SUM("NEW_TAX_SPECIAL_CURRENT_MTH")/12 IS NOT NULL))
11 - FILTER(:B1=TO_DATE(' 2018-05-31 00:00:00', 'SYYYY-MM-DD HH24:MI:SS') AND :B2='LIMIT' AND :B3='00' AND
:B4='001')
12 - FILTER("COMPANY_TYPE"='LIMIT' AND "COMPANY_TAX_TYPE"=:B1 AND "TAX_TYPE"='00' AND "TAX_MTH">=TO_DATE('
2017-03-31 00:00:00', 'SYYYY-MM-DD HH24:MI:SS') AND "TAX_MTH"<=TO_DATE(' 2018-02-28 00:00:00', 'SYYYY-MM-DD
HH24:MI:SS') AND NVL("COMPANY_GROUP",'-99999')=NVL(:B2,'-99999') AND "CITY_CODE"='001' AND "NEW_TAX_MISS_CURRENT_MTH"=TO_DATE('
2200-01-01 00:00:00', 'SYYYY-MM-DD HH24:MI:SS'))
13 - ACCESS("TAXPAYER"=:B1)
2 個解決方案
解決方案1
2 2018-07-06 10:29:52
我要嘗試的第一件事是使用 MERGE 語句而不是 UPDATE 語句 - 這讓我想到了，因為您實際上是在 where 子句中的 set 子句中重復子查詢。
我認為您的 UPDATE 可以重寫為：
MERGE INTO tax_ratio tgt
USING (SELECT city_code,
tax_type,
company_type,
taxpayer,
company_group,
company_tax_type,
SUM(new_tax_current_mth) / 12 AS ecm,
SUM(new_tax_miss_current_mth) / 12 AS esm,
SUM(new_tax_get_current_mth) / 12 AS epm,
SUM(new_tax_special_current_mth) / 12 AS ecam
FROM tax_ratio
WHERE city_code = '001'
AND company_type = 'typ_01'
AND tax_mth <= add_months(to_date('08-JUL-2015', 'dd-MON-yyyy', 'nls_date_language = english'), -3)
AND tax_mth >= add_months(to_date('08-JUL-2015', 'dd-MON-yyyy', 'nls_date_language = english'), -3)
AND eff_date = to_date('08-JUL-2015', 'dd-MON-yyyy', 'nls_date_language = english')
AND tax_type = '00'
GROUP BY city_code,
tax_type,
company_type,
taxpayer,
company_group,
company_tax_type
HAVING SUM(new_tax_current_mth) <> 0
OR SUM(new_tax_miss_current_mth) <> 0
OR SUM(new_tax_get_current_mth) <> 0
OR SUM(new_tax_special_current_mth) <> 0)) src
ON (tgt.city_code = src.city_code
AND tgt.tax_type = src.tax_type
AND tgt.company_type = src.company_type
AND tgt.taxpayer = src.taxpayer
AND NVL(tgt.company_group, '-99999') = NVL(src.company_group, '-99999')
--AND COALESCE(src.ecm, src.esm, src.epm, src.ecam) IS NOT NULL -- unnecessary, since your having clause excludes rows where all are null anyway
AND tgt.company_tax_type = src.company_tax_type)
WHEN MATCHED THEN
UPDATE SET tgt.ecm = NVL(src.ecm, 0),
tgt.esm = NVL(src.esm, 0),
tgt.epm = NVL(src.epm, 0),
tgt.ecam = NVL(src.ecam, 0),
tgt.update_dt = SYSDATE,
tgt.update_by = 'ffee_user')
WHERE tgt.tax_mth = to_date('08-JUL-2015', 'dd-MON-yyyy', 'nls_date_language = english');
筆記：
- 我更改了
to_dates()以包含格式掩碼（否則，如果您的 NLS_DATE_FORMAT nls 參數已更改，則to_date()將失敗）並且，由於您使用“JUL”作為月份，因此附加的第三個參數用於制作日期-as-a-string 轉換 nls 設置獨立（例如，如果您的 NLS_DATE_FORMAT nls 參數設置為JUL不是有效的短月，如果未設置第三個參數，則to_date()將失敗）。 您可以通過將日期作為數字傳遞來避免使用第三個參數，例如to_date('05/07/2015', 'dd/mm/yyyy')。 - 我更改了
and ecm is not null and esm is not null and...改為使用COALESCE，因為這會返回列表中的第一個非空值。 因此，您的支票變為and COALESCE(ecm, esm, ...) is not null。 - 實際上不需要 COALESCE，因為您的 HAVING 子句有效地排除了所有值為 NULL 和 0 的行。
- 用於使子查詢相關的謂詞成為目標表和源子查詢之間的連接條件，這意味着您不再需要以前在相關子查詢中的外部查詢。
- 我將條件從連接子句移動到僅更新具有特定納稅月份的行到更新部分的 WHERE 子句中。 我很確定它可以留在 join 子句中，但我認為如果將它放在更新的 WHERE 子句中，意圖會更清楚。
我會測試新的 MERGE 語句以確保它正在做正確的事情（或修復它直到它做正確）然后看看它如何影響性能。
如果它仍然作為 AWR 中的問題陳述彈出，那么我將進一步研究如何調整它； 可能需要更新/附加索引，可能需要物化視圖等。
解決方案2
0 2022-12-20 11:52:56
1.我會建議你在表級訪問大量數據的條件下創建索引。在索引上創建后。將節省時間
tax_ratio WHERE city_code ='001' AND company_type ='typ_01' AND tax_mth <= add_months(TO_DATE('08-JUL-2015'),-3) AND tax_mth >= add_months(TO_DATE('08-JUL-2015'), -14) AND eff_date =TO_DATE('08-JUL-2015') AND tax_type = '00'
用臨時表重寫復雜的子查詢。在臨時表中使用可以使用的索引來更快地檢索數據。
使用減號而不是存在的子查詢。
SQL性能調優 - 更新查詢
聲明:本站的技術帖子網頁，遵循CC BY-SA 4.0協議，如果您需要轉載，請注明本站網址或者原文地址。任何問題請咨詢:yoyou2525@163.com.