天天看点

PostgreSQL 并行计算解说 之11 - parallel gather, gather merge

标签

PostgreSQL , cpu 并行 , smp 并行 , 并行计算 , gpu 并行 , 并行过程支持

https://github.com/digoal/blog/blob/master/201903/20190317_03.md#%E8%83%8C%E6%99%AF 背景

PostgreSQL 11 优化器已经支持了非常多场合的并行。简单估计,已支持27余种场景的并行计算。

parallel seq scan                      
                      
parallel index scan                      
                      
parallel index only scan                      
                      
parallel bitmap scan                      
                      
parallel filter                      
                  
parallel hash agg                  
                  
parallel group agg                  
                      
parallel cte                      
                      
parallel subquery                      
                      
parallel create table                      
                      
parallel create index                      
                      
parallel select into                      
                      
parallel CREATE MATERIALIZED VIEW                      
                      
parallel 排序 : gather merge                       
                      
parallel nestloop join                      
                      
parallel hash join                      
                      
parallel merge join                      
                      
parallel 自定义并行聚合                      
                      
parallel 自定义并行UDF                      
                      
parallel append                      
                      
parallel union                      
                      
parallel fdw table scan                      
                      
parallel partition join                      
                      
parallel partition agg                      
                      
parallel gather              
      
parallel gather merge      
                      
parallel rc 并行                      
                      
parallel rr 并行                      
                      
parallel GPU 并行                      
                      
parallel unlogged table                       
           

接下来进行一一介绍。

关键知识请先自行了解:

1、优化器自动并行度算法 CBO

《PostgreSQL 9.6 并行计算 优化器算法浅析》 《PostgreSQL 11 并行计算算法,参数,强制并行度设置》

https://github.com/digoal/blog/blob/master/201903/20190317_03.md#parallel-gather--gather-merge parallel gather , gather merge

并行计算结果合并: (gather, gather merge)

parallel gather , gather merge 是一段并行任务中的ROOT部分,gather用来收集所有计算子进程的计算结果,gather merge用来收集所有计算子进程的计算结果并进行归并排序。

数据量:10亿。

场景 数据量 关闭并行 开启并行 并行度 开启并行性能提升倍数
普通并行(gather) 10 亿 70.2 秒 2.5 秒 30 28.1 倍
归并并行(gather merge) 78.2 秒 2.76 秒 28.3 倍

https://github.com/digoal/blog/blob/master/201903/20190317_03.md#1%E5%85%B3%E9%97%AD%E5%B9%B6%E8%A1%8C%E8%80%97%E6%97%B6-702-%E7%A7%92--782-%E7%A7%92 1、关闭并行,耗时: 70.2 秒 , 78.2 秒。

postgres=# explain select max(i) from table2;  
                                 QUERY PLAN                                   
----------------------------------------------------------------------------  
 Aggregate  (cost=16924779.00..16924779.01 rows=1 width=4)  
   ->  Seq Scan on table2  (cost=0.00..14424779.00 rows=1000000000 width=4)  
(2 rows)  
  
postgres=# explain select i from table2 order by i desc limit 10;  
                                    QUERY PLAN                                      
----------------------------------------------------------------------------------  
 Limit  (cost=10036034419.47..10036034419.50 rows=10 width=4)  
   ->  Sort  (cost=10036034419.47..10038534419.47 rows=1000000000 width=4)  
         Sort Key: i DESC  
         ->  Seq Scan on table2  (cost=0.00..14424779.00 rows=1000000000 width=4)  
(4 rows)  
  
  
postgres=# select max(i) from table2;  
 max   
-----  
  -1  
(1 row)  
  
Time: 70168.043 ms (01:10.168)  
  
postgres=# select i from table2 order by i desc limit 10;  
  i    
-----  
  -1  
  -2  
  -6  
  -7  
  -7  
 -12  
 -12  
 -14  
 -16  
 -18  
(10 rows)  
  
Time: 78217.388 ms (01:18.217)  
           

https://github.com/digoal/blog/blob/master/201903/20190317_03.md#2%E5%BC%80%E5%90%AF%E5%B9%B6%E8%A1%8C%E8%80%97%E6%97%B6-25-%E7%A7%92--276-%E7%A7%92 2、开启并行,耗时: 2.5 秒 , 2.76 秒。

postgres=# explain select max(i) from table2;  
                                          QUERY PLAN                                            
----------------------------------------------------------------------------------------------  
 Finalize Aggregate  (cost=4841445.75..4841445.76 rows=1 width=4)  
   ->  Gather  (cost=4841445.67..4841445.68 rows=30 width=4)  
         Workers Planned: 30  
         ->  Partial Aggregate  (cost=4841445.67..4841445.68 rows=1 width=4)  
               ->  Parallel Seq Scan on table2  (cost=0.00..4758112.33 rows=33333333 width=4)  
(5 rows)  
  
postgres=# explain select i from table2 order by i desc limit 10;  
                                          QUERY PLAN                                            
----------------------------------------------------------------------------------------------  
 Limit  (cost=10005478434.44..10005478434.72 rows=10 width=4)  
   ->  Gather Merge  (cost=10005478434.44..10032832749.05 rows=999999990 width=4)  
         Workers Planned: 30  
         ->  Sort  (cost=10005478433.68..10005561767.01 rows=33333333 width=4)  
               Sort Key: i DESC  
               ->  Parallel Seq Scan on table2  (cost=0.00..4758112.33 rows=33333333 width=4)  
(6 rows)  
  
  
postgres=# select max(i) from table2;  
 max   
-----  
  -1  
(1 row)  
  
Time: 2523.247 ms (00:02.523)  
  
postgres=# select i from table2 order by i desc limit 10;  
  i    
-----  
  -1  
  -2  
  -6  
  -7  
  -7  
 -12  
 -12  
 -14  
 -16  
 -18  
(10 rows)  
  
Time: 2762.977 ms (00:02.763)  
           

https://github.com/digoal/blog/blob/master/201903/20190317_03.md#%E5%85%B6%E4%BB%96%E7%9F%A5%E8%AF%86 其他知识

2、function, op 识别是否支持parallel

postgres=# select proparallel,proname from pg_proc;                      
 proparallel |                   proname                                          
-------------+----------------------------------------------                      
 s           | boolin                      
 s           | boolout                      
 s           | byteain                      
 s           | byteaout                      
           

3、subquery mapreduce unlogged table

对于一些情况,如果期望简化优化器对非常非常复杂的SQL并行优化的负担,可以自己将SQL拆成几段,中间结果使用unlogged table保存,类似mapreduce的思想。unlogged table同样支持parallel 计算。

4、vacuum,垃圾回收并行。

5、dblink 异步调用并行

《PostgreSQL VOPS 向量计算 + DBLINK异步并行 - 单实例 10亿 聚合计算跑进2秒》 《PostgreSQL 相似搜索分布式架构设计与实践 - dblink异步调用与多机并行(远程 游标+记录 UDF实例)》 《PostgreSQL dblink异步调用实现 并行hash分片JOIN - 含数据交、并、差 提速案例 - 含dblink VS pg 11 parallel hash join VS pg 11 智能分区JOIN》

暂时不允许并行的场景(将来PG会继续扩大支持范围):

1、修改行,锁行,除了create table as , select into, create mview这几个可以使用并行。

2、query 会被中断时,例如cursor , loop in PL/SQL ,因为涉及到中间处理,所以不建议开启并行。

3、paralle unsafe udf ,这种UDF不会并行

4、嵌套并行(udf (内部query并行)),外部调用这个UDF的SQL不会并行。(主要是防止large parallel workers )

5、SSI 隔离级别

https://github.com/digoal/blog/blob/master/201903/20190317_03.md#%E5%8F%82%E8%80%83 参考

https://www.postgresql.org/docs/11/parallel-plans.html 《PostgreSQL 11 preview - 并行计算 增强 汇总》 《PostgreSQL 10 自定义并行计算聚合函数的原理与实践 - (含array_agg合并多个数组为单个一元数组的例子)》

https://github.com/digoal/blog/blob/master/201903/20190317_03.md#%E5%85%8D%E8%B4%B9%E9%A2%86%E5%8F%96%E9%98%BF%E9%87%8C%E4%BA%91rds-postgresql%E5%AE%9E%E4%BE%8Becs%E8%99%9A%E6%8B%9F%E6%9C%BA 免费领取阿里云RDS PostgreSQL实例、ECS虚拟机

PostgreSQL 并行计算解说 之11 - parallel gather, gather merge