Oracle SQL消耗巨大的臨時空間

我正面臨着一個在臨時空間&查詢性能之間的折衷問題。Oracle SQL消耗巨大的臨時空間

我的情況是這樣的：我有一個巨大的事務表（跨25個分區200億條記錄）和一個有7條記錄的小型公制查詢表。我需要處理每個度量記錄的每個交易記錄。基本上輸出將是7 * 200億記錄。該輸出必須根據5-6列進行彙總。

我已經考慮了兩種選擇：

交叉連接這兩個表，並使用「案時，」指定處理邏輯WRT公制和「分組依據」操作做。
對於每個度量標識和「UNION ALL」結果有7個不同的查詢。

＃1正在消耗大約250 GB的巨大臨時空間，＃2運行大約230分鐘。

有沒有辦法，我可以優化其中的一個選項？我需要在60分鐘內完成此查詢。

添加查詢

@大衛查詢粘貼下面

WITH IDQ_LKP AS 
    (SELECT '703' AS METRIC_ID,'desc1' AS EN_METRIC_1_NM 
    FROM DUAL 
    UNION ALL 
    SELECT '702' AS METRIC_ID,'desc1' AS EN_METRIC_1_NM 
    FROM DUAL 
    UNION ALL 
    SELECT '704' AS METRIC_ID,'% desc1' AS EN_METRIC_1_NM 
    FROM DUAL 
    UNION ALL 
    SELECT '705' AS METRIC_ID,'desc2' AS EN_METRIC_1_NM 
    FROM DUAL 
    UNION ALL 
    SELECT '706' AS METRIC_ID,'desc3' AS EN_METRIC_1_NM 
    FROM DUAL 
    UNION ALL 
    SELECT '707' AS METRIC_ID,'desc5' AS EN_METRIC_1_NM 
    FROM DUAL 
    UNION ALL 
    SELECT '701' AS METRIC_ID,'desc4' AS EN_METRIC_1_NM 
    FROM DUAL) 
SELECT /*+ parallel(16) USE_HASH_AGGREGATION */ col1 , 
     col2 , 
     'Monthly Snapshots' AS Time_Rollup , col3 , date_pk , colr , col5 , colr_DESC , col5_DESC , metric_id , EN_METRIC_1_NM , sum (
    CASE 
    WHEN (metric_id='704' 
     AND record_identifier=17) 
     OR (metric_id='705' 
     AND record_identifier=123) 
     OR (metric_id='706' 
     AND record_identifier=43) 
     OR (metric_id='707' 
     AND record_identifier=34) THEN 
    nvl (record_count,0) 
    WHEN metric_id NOT IN ('704','705','706','707') THEN 
    NULL 
    ELSE 0 
    END) AS METRIC_1_CY , NULL AS METRIC_1_LY , sum (
    CASE 
    WHEN (metric_id='703' 
     AND record_identifier=17) 
     OR (metric_id='705' 
     AND record_identifier=777) 
     OR (metric_id='702' 
     AND record_identifier=123) 
     OR (metric_id='704' 
     AND record_identifier=17) 
     OR (metric_id='706' 
     AND record_identifier=99999997) 
     OR (metric_id='707' 
     AND record_identifier=99999996) THEN 
    nvl (record_count,0) 
    WHEN metric_id NOT IN ('702','703','704','705','706','707') THEN 
    NULL 
    ELSE 0 
    END) AS METRIC_2_CY , NULL AS METRIC_2_LY , NULL AS METRIC_3_CY , NULL AS METRIC_3_LY 
FROM TXN,LKP 
WHERE col1=2 
GROUP BY col1 , col2 , col3 , date_pk , colr , col5 , colr_DESC , col5_DESC, metric_id , EN_METRIC_1_NM

嗨，

我仍然看到252 GB需要臨時空間在解釋計劃...

------------------------------------------------------------------------------------------------------------------------------------------------- 
| Id | Operation      | Name      | Rows | Bytes |TempSpc| Cost | Pstart| Pstop | TQ |IN-OUT| PQ Distrib | 
------------------------------------------------------------------------------------------------------------------------------------------------- 
| 0 | SELECT STATEMENT    |       |  |  |  | 17M|  |  |  |  |   | 
| 1 | PX COORDINATOR    |       |  |  |  |  |  |  |  |  |   | 
| 2 | PX SEND QC (RANDOM)   | :TQ10003     | 1894M| 217G|  | 17M|  |  | Q1,03 | P->S | QC (RAND) | 
| 3 | HASH GROUP BY    |       | 1894M| 217G| 262G| 17M|  |  | Q1,03 | PCWP |   | 
| 4 |  PX RECEIVE     |       | 1894M| 217G|  | 17M|  |  | Q1,03 | PCWP |   | 
| 5 |  PX SEND HASH    | :TQ10002     | 1894M| 217G|  | 17M|  |  | Q1,02 | P->P | HASH  | 
| 6 |  HASH GROUP BY   |       | 1894M| 217G| 262G| 17M|  |  | Q1,02 | PCWP |   | 
| 7 |  MERGE JOIN CARTESIAN |       | 1894M| 217G|  | 149K|  |  | Q1,02 | PCWP |   | 
| 8 |   BUFFER SORT   |       |  |  |  |  |  |  | Q1,02 | PCWC |   | 
| 9 |   PX RECEIVE   |       |  7 | 154 |  | 14 |  |  | Q1,02 | PCWP |   | 
| 10 |   PX SEND BROADCAST | :TQ10000     |  7 | 154 |  | 14 |  |  |  | S->P | BROADCAST | 
| 11 |   VIEW    |       |  7 | 154 |  | 14 |  |  |  |  |   | 
| 12 |    UNION-ALL   |       |  |  |  |  |  |  |  |  |   | 
| 13 |    FAST DUAL   |       |  1 |  |  |  2 |  |  |  |  |   | 
| 14 |    FAST DUAL   |       |  1 |  |  |  2 |  |  |  |  |   | 
| 15 |    FAST DUAL   |       |  1 |  |  |  2 |  |  |  |  |   | 
| 16 |    FAST DUAL   |       |  1 |  |  |  2 |  |  |  |  |   | 
| 17 |    FAST DUAL   |       |  1 |  |  |  2 |  |  |  |  |   | 
| 18 |    FAST DUAL   |       |  1 |  |  |  2 |  |  |  |  |   | 
| 19 |    FAST DUAL   |       |  1 |  |  |  2 |  |  |  |  |   | 
| 20 |   BUFFER SORT   |       | 270M| 25G|  | 17M|  |  | Q1,02 | PCWP |   | 
| 21 |   VIEW     |       | 270M| 25G|  |  |  |  | Q1,02 | PCWP |   | 
| 22 |   HASH GROUP BY  |       | 270M| 22G| 29G| 115K|  |  | Q1,02 | PCWP |   | 
| 23 |   PX RECEIVE   |       | 270M| 22G|  | 843 |  |  | Q1,02 | PCWP |   | 
| 24 |    PX SEND HASH  | :TQ10001     | 270M| 22G|  | 843 |  |  | Q1,01 | P->P | HASH  | 
| 25 |    PX BLOCK ITERATOR |       | 270M| 22G|  | 843 | 28 | 55 | Q1,01 | PCWC |   | 
|* 26 |    TABLE ACCESS FULL| TXN      | 270M| 22G|  | 843 | 28 | 55 | Q1,01 | PCWP |   | 
---------------------------------------------------------------------------------------------------------

它正在運行...我懷疑它可能陷入同樣的問題....

計劃「UNION ALL」的方法..

------------------------------------------------------------------------------------------------------------------------------------------- 
| Id | Operation    | Name      | Rows | Bytes |TempSpc| Cost | Pstart| Pstop | TQ |IN-OUT| PQ Distrib | 
------------------------------------------------------------------------------------------------------------------------------------------- 
| 0 | SELECT STATEMENT   |       |  |  |  | 774K|  |  |  |  |   | 
| 1 | UNION-ALL    |       |  |  |  |  |  |  |  |  |   | 
| 2 | PX COORDINATOR   |       |  |  |  |  |  |  |  |  |   | 
| 3 | PX SEND QC (RANDOM) | :TQ10001     | 270M| 18G|  | 100K|  |  | Q1,01 | P->S | QC (RAND) | 
| 4 |  HASH GROUP BY  |       | 270M| 18G| 24G| 100K|  |  | Q1,01 | PCWP |   | 
| 5 |  PX RECEIVE   |       | 270M| 18G|  | 843 |  |  | Q1,01 | PCWP |   | 
| 6 |  PX SEND HASH  | :TQ10000     | 270M| 18G|  | 843 |  |  | Q1,00 | P->P | HASH  | 
| 7 |  PX BLOCK ITERATOR |       | 270M| 18G|  | 843 | 28 | 55 | Q1,00 | PCWC |   | 
|* 8 |   TABLE ACCESS FULL| TXN      | 270M| 18G|  | 843 | 28 | 55 | Q1,00 | PCWP |   | 
| 9 | PX COORDINATOR   |       |  |  |  |  |  |  |  |  |   | 
| 10 | PX SEND QC (RANDOM) | :TQ20001     | 270M| 21G|  | 112K|  |  | Q2,01 | P->S | QC (RAND) | 
| 11 |  HASH GROUP BY  |       | 270M| 21G| 28G| 112K|  |  | Q2,01 | PCWP |   | 
| 12 |  PX RECEIVE   |       | 270M| 21G|  | 843 |  |  | Q2,01 | PCWP |   | 
| 13 |  PX SEND HASH  | :TQ20000     | 270M| 21G|  | 843 |  |  | Q2,00 | P->P | HASH  | 
| 14 |  PX BLOCK ITERATOR |       | 270M| 21G|  | 843 | 28 | 55 | Q2,00 | PCWC |   | 
|* 15 |   TABLE ACCESS FULL| TXN      | 270M| 21G|  | 843 | 28 | 55 | Q2,00 | PCWP |   | 
| 16 | PX COORDINATOR   |       |  |  |  |  |  |  |  |  |   | 
| 17 | PX SEND QC (RANDOM) | :TQ30001     | 270M| 21G|  | 112K|  |  | Q3,01 | P->S | QC (RAND) | 
| 18 |  HASH GROUP BY  |       | 270M| 21G| 28G| 112K|  |  | Q3,01 | PCWP |   | 
| 19 |  PX RECEIVE   |       | 270M| 21G|  | 843 |  |  | Q3,01 | PCWP |   | 
| 20 |  PX SEND HASH  | :TQ30000     | 270M| 21G|  | 843 |  |  | Q3,00 | P->P | HASH  | 
| 21 |  PX BLOCK ITERATOR |       | 270M| 21G|  | 843 | 28 | 55 | Q3,00 | PCWC |   | 
|* 22 |   TABLE ACCESS FULL| TXN      | 270M| 21G|  | 843 | 28 | 55 | Q3,00 | PCWP |   | 
| 23 | PX COORDINATOR   |       |  |  |  |  |  |  |  |  |   | 
| 24 | PX SEND QC (RANDOM) | :TQ40001     | 270M| 21G|  | 112K|  |  | Q4,01 | P->S | QC (RAND) | 
| 25 |  HASH GROUP BY  |       | 270M| 21G| 28G| 112K|  |  | Q4,01 | PCWP |   | 
| 26 |  PX RECEIVE   |       | 270M| 21G|  | 843 |  |  | Q4,01 | PCWP |   | 
| 27 |  PX SEND HASH  | :TQ40000     | 270M| 21G|  | 843 |  |  | Q4,00 | P->P | HASH  | 
| 28 |  PX BLOCK ITERATOR |       | 270M| 21G|  | 843 | 28 | 55 | Q4,00 | PCWC |   | 
|* 29 |   TABLE ACCESS FULL| TXN      | 270M| 21G|  | 843 | 28 | 55 | Q4,00 | PCWP |   | 
| 30 | PX COORDINATOR   |       |  |  |  |  |  |  |  |  |   | 
| 31 | PX SEND QC (RANDOM) | :TQ50001     | 270M| 21G|  | 112K|  |  | Q5,01 | P->S | QC (RAND) | 
| 32 |  HASH GROUP BY  |       | 270M| 21G| 28G| 112K|  |  | Q5,01 | PCWP |   | 
| 33 |  PX RECEIVE   |       | 270M| 21G|  | 843 |  |  | Q5,01 | PCWP |   | 
| 34 |  PX SEND HASH  | :TQ50000     | 270M| 21G|  | 843 |  |  | Q5,00 | P->P | HASH  | 
| 35 |  PX BLOCK ITERATOR |       | 270M| 21G|  | 843 | 28 | 55 | Q5,00 | PCWC |   | 
|* 36 |   TABLE ACCESS FULL| TXN      | 270M| 21G|  | 843 | 28 | 55 | Q5,00 | PCWP |   | 
| 37 | PX COORDINATOR   |       |  |  |  |  |  |  |  |  |   | 
| 38 | PX SEND QC (RANDOM) | :TQ60001     | 270M| 21G|  | 112K|  |  | Q6,01 | P->S | QC (RAND) | 
| 39 |  HASH GROUP BY  |       | 270M| 21G| 28G| 112K|  |  | Q6,01 | PCWP |   | 
| 40 |  PX RECEIVE   |       | 270M| 21G|  | 843 |  |  | Q6,01 | PCWP |   | 
| 41 |  PX SEND HASH  | :TQ60000     | 270M| 21G|  | 843 |  |  | Q6,00 | P->P | HASH  | 
| 42 |  PX BLOCK ITERATOR |       | 270M| 21G|  | 843 | 28 | 55 | Q6,00 | PCWC |   | 
|* 43 |   TABLE ACCESS FULL| TXN      | 270M| 21G|  | 843 | 28 | 55 | Q6,00 | PCWP |   | 
| 44 | PX COORDINATOR   |       |  |  |  |  |  |  |  |  |   | 
| 45 | PX SEND QC (RANDOM) | :TQ70001     | 270M| 21G|  | 112K|  |  | Q7,01 | P->S | QC (RAND) | 
| 46 |  SORT GROUP BY  |       | 270M| 21G| 28G| 112K|  |  | Q7,01 | PCWP |   | 
| 47 |  PX RECEIVE   |       | 270M| 21G|  | 843 |  |  | Q7,01 | PCWP |   | 
| 48 |  PX SEND HASH  | :TQ70000     | 270M| 21G|  | 843 |  |  | Q7,00 | P->P | HASH  | 
| 49 |  PX BLOCK ITERATOR |       | 270M| 21G|  | 843 | 28 | 55 | Q7,00 | PCWC |   | 
|* 50 |   TABLE ACCESS FULL| TXN      | 270M| 21G|  | 843 | 28 | 55 | Q7,00 | PCWP |   | 
-------------------------------------------------------------------------------------------------------------------------------------------

來源

2014-02-06 user3279189

在230分鐘的運行時間裏，讀寫臨時空間的時間是多少？桌子本身佔用了多少空間？表的分區鍵列是分組列之一嗎？排序是最優的，單通還是多通？預計分組操作總共需要多少行？ –

在執行度量表連接之前是否有機會聚合事務表？ –

桌子本身佔用了多少空間？ 18 GB是表格的分區鍵也用於聚合，分組後 - 我預計接近70億條記錄......我如何發現排序是否最優？ – user3279189

看起來你可以做選擇兩個運行的七個獨立的查詢在全球不同的CPU並行。也許建立某種觀點來解決結合的需要？

來源

2014-02-06 11:06:55 unigeek

它已經並行運行...它的提取工作..so查看將不會的幫助 – user3279189

是的，我在提交後看到了Oracle標籤。想想您的硬件可能只有7個線程。 – unigeek

只看該查詢，我會考慮嘗試一種方法是從實施該方法改變：

加入大表到小表
計算度量值
聚集到要求的水平

...這個...

將大表聚合到所需級別（如果可能，應用過濾器以刪除不需要的行）。
加入到小桌子
計算指標
可能骨料再次

這應該需要更小的臨時數據集。

來源

2014-02-06 12:34:20

謝謝，我會嘗試相同的和恢復..... – user3279189

得到兩種方法的解釋計劃 –

我已經粘貼了新方法的解釋，請看看並建議...在此先感謝... – user3279189

我的評論對你的要求有點前所未聞。再說一次：爲了避免巨大的中間數據，請事先減少。請嘗試以下。首先將大表預先聚合，然後交叉連接，然後重新聚合。這減少了笛卡爾產品中的中間數據。這可能會解決您的問題。

select 
    col1, col2, 'Monthly Snapshots' AS Time_Rollup, col3, date_pk, colr, col5, colr_DESC, col5_DESC, metric_id, en_metric_1_nm 
    , sum 
    (
    case 
    when (metric_id='704' and record_identifier=17) 
    or (metric_id='705' and record_identifier=123) 
    or (metric_id='706' and record_identifier=43) 
    or (metric_id='707' and record_identifier=34) then 
     sum_record_count 
    when metric_id not in ('704','705','706','707') then 
     null 
    else 
     0 
    end 
) as metric_1_cy 
    , null as metric_1_ly 
    , sum 
    (
    case 
    when (metric_id='703' and record_identifier=17) 
    or (metric_id='705' and record_identifier=777) 
    or (metric_id='702' and record_identifier=123) 
    or (metric_id='704' and record_identifier=17) 
    or (metric_id='706' and record_identifier=99999997) 
    or (metric_id='707' and record_identifier=99999996) then 
     sum_record_count 
    when metric_id not in ('702','703','704','705','706','707') then 
     null 
    else 
     0 
    end 
) as metric_2_cy 
    , null as metric_2_ly 
    , null as metric_3_cy 
    , null as metric_3_ly 
from 
(
    select col1, col2, col3, date_pk, colr, col5, colr_desc, col5_desc, record_identifier, 
    sum(nvl(record_count,0)) as sum_record_count 
    from txn 
    where col1=2 
    group by col1, col2, col3, date_pk, colr, col5, colr_desc, col5_desc, record_identifier 
) pre_aggregate 
cross join 
(
    select '703' as metric_id,'desc1' as en_metric_1_nm from dual 
    union all 
    select '702' as metric_id,'desc1' as en_metric_1_nm from dual 
    ... 
) lkp 
group by col1, col2, col3, date_pk, colr, col5, colr_DESC, col5_DESC, metric_id, en_metric_1_nm;

來源

2014-02-06 13:31:06

謝謝......我現在正在測試它......我將回復 – user3279189

我是新來的這個論壇，有沒有一種方法可以格式化並粘貼查詢/計劃？ – user3279189

我試過這個，因爲它佔用了超過150GB的臨時空間而失敗 – user3279189

爲每個分區創建一個查詢。由於分區鍵是聚合列之一，每個分區將生成單獨的數據。

將單個語句分解爲多個較小的語句幾乎總是效率較低。但是通過分區，即使查詢了25次，整個表的數據也只會被讀取一次。這種改變應該只比原來的方法稍微低效，但應該使用少得多的臨時表空間。除非分區是可怕的傾斜。

begin 
    --Loop through all partitions. 
    for partition_names in 
    (
     select partition_name 
     from user_tab_partitions 
     where table_name = 'TXN' 
    ) loop 
     --Execute the statement. 
     execute immediate q'< 
      WITH IDQ_LKP AS 
      ... 
      FROM TXN partition (>'||partition_names.partition_name||q'<),LKP 
      ... 
     >'; 

     --Normally commits should not be in DML loops, but a commit is required 
     --if the statement uses direct-path writes. 
     commit; 
    end loop; 
end; 
/

更新 - 單查詢方法和證據表明，它可能工作

每個分區上的查詢可以合併成一個單一的SQL語句。這應該仍然表現良好，使用較少的臨時表空間，儘管它將是一個helluva大型SQL語句。

的基本思路是：

with lookup as ... 
select txn partition (partition1), lookup ... union all 
select txn partition (partition2), lookup ... union all 
... 
select txn partition (partition25), lookup ... union all

這裏的樣品臺，樣品數據和僞造統計：

create table txn(date_pk date, col1 number, col2 number, col3 number, colr number 
    ,col5 number, colr_desc number, col5_desc number, record_count number 
    ,record_identifier number) 
partition by hash (col2) 
(
    partition p01,partition p02,partition p03,partition p04,partition p05, 
    partition p06,partition p07,partition p08,partition p09,partition p10, 
    partition p11,partition p12,partition p13,partition p14,partition p15, 
    partition p16,partition p17,partition p18,partition p19,partition p20, 
    partition p21,partition p22,partition p23,partition p24,partition p25 
); 

--A few fake rows just to create a column censity for GROUP BY 
insert into txn 
select sysdate, 2, level, level, level, level, level, level, level, level 
from dual connect by level <= 100000; 

begin 
    --Gather mostly for column density. 
    --Use GLOBAL because I don't know how to fake partition stats. 
    dbms_stats.gather_table_stats(user, 'txn', granularity => 'global'); 
    --Fake 452M rows. This number will generate 262G of TempSpc. 
    --I can't reproduce it with 20G because I don't know the column densities. 
    dbms_stats.set_table_stats(user, 'TXN', numrows => 452000000); 
end; 
/

這裏的原始查詢，有一些化妝品修改，使之更小。它仍然包含相同的列，條件和分組。請注意，優化器估計它將使用262GB的臨時表空間，並且執行與您的執行非常相似。

explain plan for 
WITH IDQ_LKP AS 
(
    SELECT '703' AS METRIC_ID,'desc1' AS EN_METRIC_1_NM FROM DUAL UNION ALL 
    SELECT '702' AS METRIC_ID,'desc1' AS EN_METRIC_1_NM FROM DUAL UNION ALL 
    SELECT '704' AS METRIC_ID,'% desc1' AS EN_METRIC_1_NM FROM DUAL UNION ALL 
    SELECT '705' AS METRIC_ID,'desc2' AS EN_METRIC_1_NM FROM DUAL UNION ALL 
    SELECT '706' AS METRIC_ID,'desc3' AS EN_METRIC_1_NM FROM DUAL UNION ALL 
    SELECT '707' AS METRIC_ID,'desc5' AS EN_METRIC_1_NM FROM DUAL UNION ALL 
    SELECT '701' AS METRIC_ID,'desc4' AS EN_METRIC_1_NM FROM DUAL 
) 
SELECT /*+ parallel(16) */ 
    col1,col2,'Monthly Snapshots' Time_Rollup,col3,date_pk,colr,col5,colr_DESC,col5_DESC,metric_id,EN_METRIC_1_NM,NULL METRIC_1_LY,NULL METRIC_2_LY,NULL METRIC_3_CY,NULL METRIC_3_LY 
    ,sum (CASE WHEN (metric_id='704' AND record_identifier=17) THEN nvl (record_count,0) ELSE 0 END) AS METRIC_1_CY 
    ,sum (CASE WHEN (metric_id='703' AND record_identifier=17) THEN nvl (record_count,0) ELSE 0 END) AS METRIC_2_CY 
FROM TXN,IDQ_LKP 
WHERE col1=2 
GROUP BY col1 , col2 , col3 , date_pk , colr , col5 , colr_DESC , col5_DESC, metric_id , EN_METRIC_1_NM; 

select * from table(dbms_xplan.display); 


Plan hash value: 2764457837 

-------------------------------------------------------------------------------------------------------------------------------------------- 
| Id | Operation     | Name  | Rows | Bytes |TempSpc| Cost (%CPU)| Time  | Pstart| Pstop | TQ |IN-OUT| PQ Distrib | 
-------------------------------------------------------------------------------------------------------------------------------------------- 
| 0 | SELECT STATEMENT   |   | 3163M| 179G|  | 67M (1)| 00:43:46 |  |  |  |  |   | 
| 1 | PX COORDINATOR   |   |  |  |  |   |   |  |  |  |  |   | 
| 2 | PX SEND QC (RANDOM)  | :TQ10002 | 3163M| 179G|  | 67M (1)| 00:43:46 |  |  | Q1,02 | P->S | QC (RAND) | 
| 3 | HASH GROUP BY   |   | 3163M| 179G| 262G| 67M (1)| 00:43:46 |  |  | Q1,02 | PCWP |   | 
| 4 |  PX RECEIVE    |   | 3163M| 179G|  | 67M (1)| 00:43:46 |  |  | Q1,02 | PCWP |   | 
| 5 |  PX SEND HASH   | :TQ10001 | 3163M| 179G|  | 67M (1)| 00:43:46 |  |  | Q1,01 | P->P | HASH  | 
| 6 |  HASH GROUP BY  |   | 3163M| 179G| 262G| 67M (1)| 00:43:46 |  |  | Q1,01 | PCWP |   | 
| 7 |  MERGE JOIN CARTESIAN|   | 3163M| 179G|  | 488K (1)| 00:00:20 |  |  | Q1,01 | PCWP |   | 
| 8 |   PX RECEIVE   |   |  7 | 70 |  | 14 (0)| 00:00:01 |  |  | Q1,01 | PCWP |   | 
| 9 |   PX SEND BROADCAST | :TQ10000 |  7 | 70 |  | 14 (0)| 00:00:01 |  |  | Q1,00 | S->P | BROADCAST | 
| 10 |   PX SELECTOR  |   |  |  |  |   |   |  |  | Q1,00 | SCWC |   | 
| 11 |   VIEW   |   |  7 | 70 |  | 14 (0)| 00:00:01 |  |  | Q1,00 | SCWC |   | 
| 12 |    UNION-ALL  |   |  |  |  |   |   |  |  | Q1,00 | SCWC |   | 
| 13 |    FAST DUAL  |   |  1 |  |  |  2 (0)| 00:00:01 |  |  | Q1,00 | SCWP |   | 
| 14 |    FAST DUAL  |   |  1 |  |  |  2 (0)| 00:00:01 |  |  | Q1,00 | SCWP |   | 
| 15 |    FAST DUAL  |   |  1 |  |  |  2 (0)| 00:00:01 |  |  | Q1,00 | SCWP |   | 
| 16 |    FAST DUAL  |   |  1 |  |  |  2 (0)| 00:00:01 |  |  | Q1,00 | SCWP |   | 
| 17 |    FAST DUAL  |   |  1 |  |  |  2 (0)| 00:00:01 |  |  | Q1,00 | SCWP |   | 
| 18 |    FAST DUAL  |   |  1 |  |  |  2 (0)| 00:00:01 |  |  | Q1,00 | SCWP |   | 
| 19 |    FAST DUAL  |   |  1 |  |  |  2 (0)| 00:00:01 |  |  | Q1,00 | SCWP |   | 
| 20 |   BUFFER SORT  |   | 451M| 21G|  | 67M (1)| 00:43:46 |  |  | Q1,01 | PCWP |   | 
| 21 |   PX BLOCK ITERATOR |   | 451M| 21G|  | 570 (95)| 00:00:01 |  1 | 25 | Q1,01 | PCWC |   | 
|* 22 |   TABLE ACCESS FULL| TXN  | 451M| 21G|  | 570 (95)| 00:00:01 |  1 | 25 | Q1,01 | PCWP |   | 
-------------------------------------------------------------------------------------------------------------------------------------------- 

Predicate Information (identified by operation id): 
--------------------------------------------------- 

    22 - filter("COL1"=2) 

Note 
----- 
    - Degree of Parallelism is 16 because of hint

更換單txn,idq_lkp具有多個txn partition (pXX),idq_lkp降低最大臨時表估計從262G到10G。此示例僅包含25個分區中的2個，但添加更多分區不會增加所需的臨時表空間。

explain plan for 
WITH IDQ_LKP AS 
(
    SELECT '703' AS METRIC_ID,'desc1' AS EN_METRIC_1_NM FROM DUAL UNION ALL 
    SELECT '702' AS METRIC_ID,'desc1' AS EN_METRIC_1_NM FROM DUAL UNION ALL 
    SELECT '704' AS METRIC_ID,'% desc1' AS EN_METRIC_1_NM FROM DUAL UNION ALL 
    SELECT '705' AS METRIC_ID,'desc2' AS EN_METRIC_1_NM FROM DUAL UNION ALL 
    SELECT '706' AS METRIC_ID,'desc3' AS EN_METRIC_1_NM FROM DUAL UNION ALL 
    SELECT '707' AS METRIC_ID,'desc5' AS EN_METRIC_1_NM FROM DUAL UNION ALL 
    SELECT '701' AS METRIC_ID,'desc4' AS EN_METRIC_1_NM FROM DUAL 
) 
SELECT /*+ parallel(16) */ 
    col1,col2,'Monthly Snapshots' Time_Rollup,col3,date_pk,colr,col5,colr_DESC,col5_DESC,metric_id,EN_METRIC_1_NM,NULL METRIC_1_LY,NULL METRIC_2_LY,NULL METRIC_3_CY,NULL METRIC_3_LY 
    ,sum (CASE WHEN (metric_id='704' AND record_identifier=17) THEN nvl (record_count,0) ELSE 0 END) AS METRIC_1_CY 
    ,sum (CASE WHEN (metric_id='703' AND record_identifier=17) THEN nvl (record_count,0) ELSE 0 END) AS METRIC_2_CY 
FROM TXN partition (p01),IDQ_LKP 
WHERE col1=2 
GROUP BY col1 , col2 , col3 , date_pk , colr , col5 , colr_DESC , col5_DESC, metric_id , EN_METRIC_1_NM 
union all 
SELECT /*+ parallel(16) */ 
    col1,col2,'Monthly Snapshots' Time_Rollup,col3,date_pk,colr,col5,colr_DESC,col5_DESC,metric_id,EN_METRIC_1_NM,NULL METRIC_1_LY,NULL METRIC_2_LY,NULL METRIC_3_CY,NULL METRIC_3_LY 
    ,sum (CASE WHEN (metric_id='704' AND record_identifier=17) THEN nvl (record_count,0) ELSE 0 END) AS METRIC_1_CY 
    ,sum (CASE WHEN (metric_id='703' AND record_identifier=17) THEN nvl (record_count,0) ELSE 0 END) AS METRIC_2_CY 
FROM TXN partition (p02),IDQ_LKP 
WHERE col1=2 
GROUP BY col1 , col2 , col3 , date_pk , colr , col5 , colr_DESC , col5_DESC, metric_id , EN_METRIC_1_NM; 


select * from table(dbms_xplan.display); 

Plan hash value: 260310120 

----------------------------------------------------------------------------------------------------------------------------------------------------------------- 
| Id | Operation      | Name      | Rows | Bytes |TempSpc| Cost (%CPU)| Time  | Pstart| Pstop | TQ |IN-OUT| PQ Distrib | 
----------------------------------------------------------------------------------------------------------------------------------------------------------------- 
| 0 | SELECT STATEMENT    |       | 253M| 14G|  | 3923K (51)| 00:02:34 |  |  |  |  |   | 
| 1 | TEMP TABLE TRANSFORMATION  |       |  |  |  |   |   |  |  |  |  |   | 
| 2 | LOAD AS SELECT    | SYS_TEMP_0FD9D66A0_D66798 |  |  |  |   |   |  |  |  |  |   | 
| 3 | UNION-ALL     |       |  |  |  |   |   |  |  |  |  |   | 
| 4 |  FAST DUAL     |       |  1 |  |  |  2 (0)| 00:00:01 |  |  |  |  |   | 
| 5 |  FAST DUAL     |       |  1 |  |  |  2 (0)| 00:00:01 |  |  |  |  |   | 
| 6 |  FAST DUAL     |       |  1 |  |  |  2 (0)| 00:00:01 |  |  |  |  |   | 
| 7 |  FAST DUAL     |       |  1 |  |  |  2 (0)| 00:00:01 |  |  |  |  |   | 
| 8 |  FAST DUAL     |       |  1 |  |  |  2 (0)| 00:00:01 |  |  |  |  |   | 
| 9 |  FAST DUAL     |       |  1 |  |  |  2 (0)| 00:00:01 |  |  |  |  |   | 
| 10 |  FAST DUAL     |       |  1 |  |  |  2 (0)| 00:00:01 |  |  |  |  |   | 
| 11 | UNION-ALL     |       |  |  |  |   |   |  |  |  |  |   | 
| 12 | PX COORDINATOR    |       |  |  |  |   |   |  |  |  |  |   | 
| 13 |  PX SEND QC (RANDOM)  | :TQ10002     | 126M| 7362M|  | 1961K (1)| 00:01:17 |  |  | Q1,02 | P->S | QC (RAND) | 
| 14 |  HASH GROUP BY    |       | 126M| 7362M| 10G| 1961K (1)| 00:01:17 |  |  | Q1,02 | PCWP |   | 
| 15 |  PX RECEIVE    |       | 126M| 7362M|  | 1961K (1)| 00:01:17 |  |  | Q1,02 | PCWP |   | 
| 16 |  PX SEND HASH   | :TQ10001     | 126M| 7362M|  | 1961K (1)| 00:01:17 |  |  | Q1,01 | P->P | HASH  | 
| 17 |   HASH GROUP BY   |       | 126M| 7362M| 10G| 1961K (1)| 00:01:17 |  |  | Q1,01 | PCWP |   | 
| 18 |   MERGE JOIN CARTESIAN |       | 126M| 7362M|  | 19567 (1)| 00:00:01 |  |  | Q1,01 | PCWP |   | 
| 19 |   PX RECEIVE   |       |  7 | 70 |  |  2 (0)| 00:00:01 |  |  | Q1,01 | PCWP |   | 
| 20 |   PX SEND BROADCAST | :TQ10000     |  7 | 70 |  |  2 (0)| 00:00:01 |  |  | Q1,00 | P->P | BROADCAST | 
| 21 |    VIEW    |       |  7 | 70 |  |  2 (0)| 00:00:01 |  |  | Q1,00 | PCWP |   | 
| 22 |    PX BLOCK ITERATOR |       |  7 | 70 |  |  2 (0)| 00:00:01 |  |  | Q1,00 | PCWC |   | 
| 23 |    TABLE ACCESS FULL| SYS_TEMP_0FD9D66A0_D66798 |  7 | 70 |  |  2 (0)| 00:00:01 |  |  | Q1,00 | PCWP |   | 
| 24 |   BUFFER SORT   |       | 18M| 879M|  | 1961K (1)| 00:01:17 |  |  | Q1,01 | PCWP |   | 
| 25 |   PX BLOCK ITERATOR |       | 18M| 879M|  | 23 (92)| 00:00:01 |  1 |  1 | Q1,01 | PCWC |   | 
|* 26 |    TABLE ACCESS FULL | TXN      | 18M| 879M|  | 23 (92)| 00:00:01 |  1 |  1 | Q1,01 | PCWP |   | 
| 27 | PX COORDINATOR    |       |  |  |  |   |   |  |  |  |  |   | 
| 28 |  PX SEND QC (RANDOM)  | :TQ20002     | 126M| 7362M|  | 1961K (1)| 00:01:17 |  |  | Q2,02 | P->S | QC (RAND) | 
| 29 |  HASH GROUP BY    |       | 126M| 7362M| 10G| 1961K (1)| 00:01:17 |  |  | Q2,02 | PCWP |   | 
| 30 |  PX RECEIVE    |       | 126M| 7362M|  | 1961K (1)| 00:01:17 |  |  | Q2,02 | PCWP |   | 
| 31 |  PX SEND HASH   | :TQ20001     | 126M| 7362M|  | 1961K (1)| 00:01:17 |  |  | Q2,01 | P->P | HASH  | 
| 32 |   HASH GROUP BY   |       | 126M| 7362M| 10G| 1961K (1)| 00:01:17 |  |  | Q2,01 | PCWP |   | 
| 33 |   MERGE JOIN CARTESIAN |       | 126M| 7362M|  | 19567 (1)| 00:00:01 |  |  | Q2,01 | PCWP |   | 
| 34 |   PX RECEIVE   |       |  7 | 70 |  |  2 (0)| 00:00:01 |  |  | Q2,01 | PCWP |   | 
| 35 |   PX SEND BROADCAST | :TQ20000     |  7 | 70 |  |  2 (0)| 00:00:01 |  |  | Q2,00 | P->P | BROADCAST | 
| 36 |    VIEW    |       |  7 | 70 |  |  2 (0)| 00:00:01 |  |  | Q2,00 | PCWP |   | 
| 37 |    PX BLOCK ITERATOR |       |  7 | 70 |  |  2 (0)| 00:00:01 |  |  | Q2,00 | PCWC |   | 
| 38 |    TABLE ACCESS FULL| SYS_TEMP_0FD9D66A0_D66798 |  7 | 70 |  |  2 (0)| 00:00:01 |  |  | Q2,00 | PCWP |   | 
| 39 |   BUFFER SORT   |       | 18M| 879M|  | 1961K (1)| 00:01:17 |  |  | Q2,01 | PCWP |   | 
| 40 |   PX BLOCK ITERATOR |       | 18M| 879M|  | 23 (92)| 00:00:01 |  2 |  2 | Q2,01 | PCWC |   | 
|* 41 |    TABLE ACCESS FULL | TXN      | 18M| 879M|  | 23 (92)| 00:00:01 |  2 |  2 | Q2,01 | PCWP |   | 
----------------------------------------------------------------------------------------------------------------------------------------------------------------- 

Predicate Information (identified by operation id): 
--------------------------------------------------- 

    26 - filter("COL1"=2) 
    41 - filter("COL1"=2) 

Note 
----- 
    - Degree of Parallelism is 16 because of hint

來源

2014-02-06 19:03:13

Sory，我必須在純SQL中實現這個功能....它是我們內部工具的一個限制... – user3279189

您可以通過25個不同查詢中的「UNION ALL」來完成此操作。這將是一個非常大的查詢，但我認爲它會很快並且使用較少的表空間。我會盡量在明天發佈更多細節。 –

嗨Joanearles，謝謝....我會嘗試我的方案，並回復給你...後續問題 - 我原來的查詢是能夠實現分區修剪...是相當於查詢單個分區在「 UNION ALL「模式？ – user3279189

Oracle SQL消耗巨大的臨時空間

回答

相關問題