PostgreSQL 11 新特性解讀: 分割槽表增加雜湊分割槽
PostgreSQL 11 的一個重量級新特性為分割槽表得到較大增強,例如支援雜湊分割槽(HASH)表,因此 PostgreSQL 支援範圍分割槽(RANGE)、列表分割槽(LIST)、>雜湊分割槽(HASH)三種分割槽方式,本文簡單演示下雜湊分割槽表。
Hash Partitioning
The table is partitioned by specifying a modulus and a remainder for each partition. Each partition will hold the rows for which the hash value of the partition key divided by the specified modulus will produce the specified remainder.
Hash分割槽表的分割槽定義包含兩個屬性,如下:
- modulus: 指Hash分割槽個數。
- remainder: 指Hash分割槽鍵取模餘。
建立分割槽表語法
CREATE TABLE table_name ( ... )
[ PARTITION BY { RANGE | LIST | HASH } ( { column_name | ( expression ) }
CREATE TABLE table_name
PARTITION OF parent_table [ (
) ] FOR VALUES partition_bound_spec
建立資料生成函式
為了便於生成測試資料,建立以下兩個函式用來隨機生成指定長度的字串,建立 random_range(int4, int4) 函式如下:
CREATE OR REPLACE FUNCTION random_range(int4, int4)
RETURNS int4
LANGUAGE SQL
AS
$$
SELECT ($1 + FLOOR(($2 - $1 + 1) * random() ))::int4;
$$
;
接著建立random_text_simple(length int4)函式,此函式會呼叫random_range(int4, int4)函式。
CREATE OR REPLACE FUNCTION random_text_simple(length int4) RETURNS text LANGUAGE PLPGSQL AS $$ DECLARE possible_chars text := '0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ'; output text := ''; i int4; pos int4; BEGIN FOR i IN 1..length LOOP pos := random_range(1, length(possible_chars)); output := output || substr(possible_chars, pos, 1); END LOOP; RETURN output; END; $$ ;
random_text_simple(length int4)函式可以隨機生成指定長度字串,如下隨機生成含三位字元的字串。
mydb=> SELECT random_text_simple(3);
random_text_simple
--------------------
LL9
(1 row)
隨機生成含六位字元的字串,如下所示:
mydb=> SELECT random_text_simple(6);
random_text_simple
--------------------
B81BPW
(1 row)
後面會用到這個函式生成測試資料。
建立雜湊分割槽父表
CREATE TABLE student (
stuname text ,
ctime timestamp(6) without time zone
) PARTITION BY HASH(stuname);
建立索引
CREATE INDEX idx_stuendt_stuname on student using btree(stuname);
建立子表
CREATE TABLE student_p0 PARTITION OF student FOR VALUES WITH(MODULUS 4, REMAINDER 0);
CREATE TABLE student_p1 PARTITION OF student FOR VALUES WITH(MODULUS 4, REMAINDER 1);
CREATE TABLE student_p2 PARTITION OF student FOR VALUES WITH(MODULUS 4, REMAINDER 2);
CREATE TABLE student_p3 PARTITION OF student FOR VALUES WITH(MODULUS 4, REMAINDER 3);
檢視分割槽表定義
francs=> \d+ student
Table "francs.student"
Column | Type | Collation | Nullable | Default | Storage | Stats target | Description
---------+--------------------------------+-----------+----------+---------+----------+--------------+-------------
stuname | text | | | | extended | |
ctime | timestamp(6) without time zone | | | | plain | |
Partition key: HASH (stuname)
Indexes:
"idx_stuendt_stuname" btree (stuname)
Partitions: student_p0 FOR VALUES WITH (modulus 4, remainder 0),
student_p1 FOR VALUES WITH (modulus 4, remainder 1),
student_p2 FOR VALUES WITH (modulus 4, remainder 2),
student_p3 FOR VALUES WITH (modulus 4, remainder 3)
從以上看出表 student 和它的四個分割槽。
插入測試資料
使用之前建立的函式 random_text_simple() 生成100萬測試資料,如下。
INSERT INTO student(stuname,ctime) SELECT random_text_simple(6),clock_timestamp() FROM generate_series(1,1000000);
檢視分割槽表資料
表資料如下
francs=> SELECT * FROM student LIMIT 3;
stuname | ctime
---------+---------------------
4JJOPN | 2018-09-20 10:45:06
NHQONC | 2018-09-20 10:45:06
8V5BGH | 2018-09-20 10:45:06
(3 rows)
統計分割槽資料量
francs=> SELECT tableoid::regclass,count(*) from student group by 1 order by 1;
tableoid | count
------------+--------
student_p0 | 250510
student_p1 | 249448
student_p2 | 249620
student_p3 | 250422
(4 rows)
可見資料均勻分佈到了四個分割槽。
根據分割槽鍵查詢
francs=> EXPLAIN ANALYZE SELECT * FROM student WHERE stuname='3LXBEV';
QUERY PLAN
------------------------------------------------------------------------------------------------------------------------------------
Append (cost=0.42..8.44 rows=1 width=15) (actual time=0.017..0.018 rows=1 loops=1)
-> Index Scan using student_p3_stuname_idx on student_p3 (cost=0.42..8.44 rows=1 width=15) (actual time=0.017..0.017 rows=1 loops=1)
Index Cond: (stuname = '3LXBEV'::text)
Planning Time: 0.198 ms
Execution Time: 0.042 ms
(5 rows)
根據分割槽鍵stuname查詢僅掃描分割槽 student_p3,並走了索引。
根據非分割槽鍵查詢
francs=> EXPLAIN ANALYZE SELECT * FROM student WHERE ctime='2018-09-20 10:53:55.48392';
QUERY PLAN
-------------------------------------------------------------------------------------------------------------------------------
Gather (cost=1000.00..13761.36 rows=4 width=15) (actual time=37.891..39.183 rows=1 loops=1)
Workers Planned: 2
Workers Launched: 2
-> Parallel Append (cost=0.00..12760.96 rows=4 width=15) (actual time=23.753..35.006 rows=0 loops=3)
-> Parallel Seq Scan on student_p0 (cost=0.00..3196.99 rows=1 width=15) (actual time=0.014..28.550 rows=1 loops=1)
Filter: (ctime = '2018-09-20 10:53:55.48392'::timestamp without time zone)
Rows Removed by Filter: 250509
-> Parallel Seq Scan on student_p3 (cost=0.00..3195.34 rows=1 width=15) (actual time=29.543..29.543 rows=0 loops=1)
Filter: (ctime = '2018-09-20 10:53:55.48392'::timestamp without time zone)
Rows Removed by Filter: 250422
-> Parallel Seq Scan on student_p2 (cost=0.00..3185.44 rows=1 width=15) (actual time=8.260..8.260 rows=0 loops=3)
Filter: (ctime = '2018-09-20 10:53:55.48392'::timestamp without time zone)
Rows Removed by Filter: 83207
-> Parallel Seq Scan on student_p1 (cost=0.00..3183.18 rows=1 width=15) (actual time=22.135..22.135 rows=0 loops=1)
Filter: (ctime = '2018-09-20 10:53:55.48392'::timestamp without time zone)
Rows Removed by Filter: 249448
Planning Time: 0.183 ms
Execution Time: 39.219 ms
(18 rows)
根據非分割槽鍵ctime查詢掃描了分割槽表所有分割槽。
總結
本文演示了 PostgreSQL 雜湊分割槽表的建立、測試資料的生成匯入和查詢計劃,後面部落格演示分割槽表增強的其它方面。
參考
- CREATE TABLE
- Table Partitioning
- WAITING FOR POSTGRESQL 11 – ADD HASH PARTITIONING
- PostgreSQL11: 分割槽表增加雜湊分割槽
- PostgreSQL11: 分割槽表支援建立主鍵、外來鍵、索引
- PostgreSQL11: 分割槽表支援UPDATE分割槽鍵
- PostgreSQL11: 分割槽表增加 Default Partition
- PostgreSQL11: Partitioning Improvements
新書推薦
最後推薦和張文升共同編寫的《PostgreSQL實戰》,本書基於PostgreSQL 10 編寫,共18章,重點介紹SQL高階特性、並行查詢、分割槽表、物理複製、邏輯複製、備份恢復、高可用、效能優化、PostGIS等,涵蓋大量實戰用例!