PLSQL_性能優化系列14_Oracle High Water Level高水位分析

2014-10-04 Created By BaoXinjian

1、摘要

---

PLSQL_性能優化系列14_Oracle High Water Level高水位分析

高水位線比如水庫中儲水的水位線，用於描述數據庫中段的擴展方式。高水位線對全表掃描方式有着相當重要的影響。

當使用delete 操做表記錄時，高水位線並不會降低，隨之致使的是全表掃描的實際開銷並無任何減小。

本文給出高水位線的描述，如何下降高水位線，以及高水位線對全表掃描的影響。

 

1. 何謂高水位線

如前所述，相似於水庫中儲水的水位線。只不過在數據庫中用於描述段的擴展方式。

能夠將數據段或索引段等想象爲一個從左到右依次排開的一系列塊。當這些塊中未填充任何數據時，高水位線位於塊的最左端(底端)

隨着記錄的不斷增長，新塊不斷地被填充並使用，高水位線隨之向右移動。高水位線之上爲未格式化的數據塊。

刪除(delete)操做以後，高水位線之下的塊處於空閒狀態，但高水位線並不隨之降低，直到重建，截斷或收縮表段。

全表掃描會掃描高水位線之下的全部塊，包括空閒數據塊(執行了delete操做)。

2. 低高水位線

是在使用ASSM時的一個概念。即便用ASSM時除了高水位線以外，還包括一個低高水位線。低高水位線必定是位於高水位線之下。

當段使用MSSM管理方式時只有一種狀況即只存在一個高水位線。

使用MMSM時，當HWM升高時，Oracle當即格式化全部塊且有效，並能夠安全讀取。僅當第一次使用時完成格式化，便於安全讀取數據。

使用ASSM時，當HWM升高時，Oracle並不會當即格式化全部塊。僅當第一次使用時完成格式化，便於安全讀取數據。

使用低高水位線能夠減小當全面掃描表段時，低高水位線與高水位線之間不安全塊的檢查數量。即低高水位線之下的塊再也不檢查。

 

2、案例 - Delete / SHRINK SPACE CASCADE / Truncate Table 對水位線的影響

---

1. 建立測試表和資料，並分析

Step1. 建立表

CREATE TABLE sh.bxj_high_water_level AS
       SELECT   ROWNUM AS id, ROUND (DBMS_RANDOM.normal * 1000) AS val1, DBMS_RANDOM.string ('p', 250) AS pad FROM DUAL CONNECT BY   LEVEL <= 10000;

Step2. 收集表的統計信息

BEGIN DBMS_STATS.gather_table_stats ('SH', 'BXJ_HIGH_WATER_LEVEL', cascade   => TRUE); END; ANALYZE TABLE sh.bxj_high_water_level COMPUTE STATISTICS;

Step3. 表的統計信息和Block信息

 

SQL> select count(*) from sh.bxj_high_water_level; Execution Plan
----------------------------------------------------------
Plan hash value: 4214873579

-----------------------------------------------------------------------------------
| Id  | Operation          | Name                 | Rows  | Cost (%CPU)| Time     |
-----------------------------------------------------------------------------------
|   0 | SELECT STATEMENT   |                      |     1 |   107   (0)| 00:00:02 |
|   1 |  SORT AGGREGATE    |                      |     1 |            |          |
|   2 |   TABLE ACCESS FULL| BXJ_HIGH_WATER_LEVEL | 10000 |   107   (0)| 00:00:02 |
-----------------------------------------------------------------------------------


Statistics
----------------------------------------------------------
          1 recursive calls 0 db block gets 375 consistent gets 0 physical reads 0 redo size 422  bytes sent via SQL*Net to client 419  bytes received via SQL*Net from client 2  SQL*Net roundtrips to/from client 0 sorts (memory) 0  sorts (disk) 1  rows processed

 

2. Delete Record 對水位線的影響

Step1. 刪除表中記錄

DELETE FROM sh.bxj_high_water_level WHERE ROWNUM <= 9900;

Step2. 收集表的統計信息 

BEGIN DBMS_STATS.gather_table_stats ('SH', 'BXJ_HIGH_WATER_LEVEL', cascade   => TRUE); END; ANALYZE TABLE sh.bxj_high_water_level COMPUTE STATISTICS;

Step3. 表的統計信息和Block信息 

SQL> select count(*) from sh.bxj_high_water_level; Execution Plan
----------------------------------------------------------
Plan hash value: 4214873579

-----------------------------------------------------------------------------------
| Id  | Operation          | Name                 | Rows  | Cost (%CPU)| Time     |
-----------------------------------------------------------------------------------
|   0 | SELECT STATEMENT   |                      |     1 |   107   (0)| 00:00:02 |
|   1 |  SORT AGGREGATE    |                      |     1 |            |          |
|   2 |   TABLE ACCESS FULL| BXJ_HIGH_WATER_LEVEL |   100 |   107   (0)| 00:00:02 |
-----------------------------------------------------------------------------------


Statistics
----------------------------------------------------------
          1 recursive calls 0 db block gets 375 consistent gets 0 physical reads 0 redo size 422  bytes sent via SQL*Net to client 419  bytes received via SQL*Net from client 2  SQL*Net roundtrips to/from client 0 sorts (memory) 0  sorts (disk) 1  rows processed

 

3.  SHRINK SPACE CASCADE 對水位線的影響

Step1。合併控件

ALTER TABLE sh.bxj_high_water_level ENABLE ROW MOVEMENT; ALTER TABLE sh.bxj_high_water_level SHRINK SPACE CASCADE; ALTER TABLE sh.bxj_high_water_level DISABLE ROW MOVEMENT; 

Step2. 收集表的統計信息  

BEGIN DBMS_STATS.gather_table_stats ('SH', 'BXJ_HIGH_WATER_LEVEL', cascade   => TRUE); END; ANALYZE TABLE sh.bxj_high_water_level COMPUTE STATISTICS; 

Step3. 表的統計信息和Block信息  

SQL> select count(*) from sh.bxj_high_water_level; Execution Plan
----------------------------------------------------------
Plan hash value: 4214873579

-----------------------------------------------------------------------------------
| Id  | Operation          | Name                 | Rows  | Cost (%CPU)| Time     |
-----------------------------------------------------------------------------------
|   0 | SELECT STATEMENT   |                      |     1 |     3   (0)| 00:00:01 |
|   1 |  SORT AGGREGATE    |                      |     1 |            |          |
|   2 |   TABLE ACCESS FULL| BXJ_HIGH_WATER_LEVEL |   100 |     3   (0)| 00:00:01 |
-----------------------------------------------------------------------------------


Statistics
----------------------------------------------------------
          1 recursive calls 0 db block gets 6 consistent gets 0 physical reads 0 redo size 422  bytes sent via SQL*Net to client 419  bytes received via SQL*Net from client 2  SQL*Net roundtrips to/from client 0 sorts (memory) 0  sorts (disk) 1  rows processed

 

4. Truncate Table 對水位線的影響

Step1. Truncate 表

TRUNCATE TABLE sh.bxj_high_water_level

Step2. 收集表的統計信息

BEGIN DBMS_STATS.gather_table_stats ('SH', 'BXJ_HIGH_WATER_LEVEL', cascade   => TRUE); END; ANALYZE TABLE bxj_water_level COMPUTE STATISTICS;

Step3. 表的統計信息和Block信息 

SQL>  select count(*) from sh.bxj_high_water_level; Execution Plan
----------------------------------------------------------
Plan hash value: 4214873579

-----------------------------------------------------------------------------------
| Id  | Operation          | Name                 | Rows  | Cost (%CPU)| Time     |
-----------------------------------------------------------------------------------
|   0 | SELECT STATEMENT   |                      |     1 |     2   (0)| 00:00:01 |
|   1 |  SORT AGGREGATE    |                      |     1 |            |          |
|   2 |   TABLE ACCESS FULL| BXJ_HIGH_WATER_LEVEL |     1 |     2   (0)| 00:00:01 |
-----------------------------------------------------------------------------------


Statistics
----------------------------------------------------------
          1 recursive calls 0 db block gets 3 consistent gets 0 physical reads 0 redo size 421  bytes sent via SQL*Net to client 419  bytes received via SQL*Net from client 2  SQL*Net roundtrips to/from client 0 sorts (memory) 0  sorts (disk) 1  rows processed

 

6、案例 - 總結

---

(1). 高水線直接決定了全表掃描所須要的I/O開銷

(2). delete操做不會下降高水位線，高水位線之下的全部塊依然被掃描

(3). 使用truncate 會重置高水位線到0位

(4). 按期使用alter table tab_name shrink space cascade 有效減小該對象上的I/O開銷

 

Thanks and Regards

參考：了沙彌 http://blog.csdn.net/leshami/article/details/6949179