熟悉oracle 的人都知道，对于两表的关联更新，其执行计划主要有 Filter 和 Outer Join 两种方式。对于大批量数据的update，Join方式明显是更优的选择。KingbaseES 也支持两种方式的关联update，语法上采用两种不同的写法。
以下以例子的形式展示两种写法及性能上的差异。这些例子同时通过KingbaseES V8R6环境验证。
一、准备测试数据
1 2 3 4 5 6 7 8 9 10 11 | create table t1(id1 integer,name1 varchar(200));
create table t2(id2 integer,name2 varchar(200));
insert into t1 select * from (select generate_series(1,1000000),repeat('a',50)) as a order by random();
insert into t2 select * from (select generate_series(1,1000000),repeat('b',50)) as a order by random();
create index ind_t1_id1 on t1(id1);
create index ind_t2_id2 on t2(id2);
analyze t1;
analyze t2;
|
二、性能测试
1、语法一
采用类似oracle filter 方式，逐条处理t1 表的每条记录。对于t1表的每条记录，都需要访问t2表。
1 2 3 4 5 6 7 8 9 10 11 | test=# explain analyze update t1 set name1=(select name2 from t2 where id1=id2);
QUERY PLAN
------------------------------------------------------------------------------------------------------------------------------------
Update on t1 (cost=0.00..8462810.00 rows=1000000 width=428) (actual time=13072.720..13072.721 rows=0 loops=1)
-> Seq Scan on t1 (cost=0.00..8462810.00 rows=1000000 width=428) (actual time=0.035..6620.732 rows=1000000 loops=1)
SubPlan 1
-> Index Scan using ind_t2_id2 on t2 (cost=0.42..8.44 rows=1 width=51) (actual time=0.006..0.006 rows=1 loops=1000000)
Index Cond: (id2 = t1.id1)
Planning Time: 0.116 ms
Execution Time: 13072.780 ms
(7 rows)
|
2、语法二
采用hash join，大批量的update 效率更高。
1 2 3 4 5 6 7 8 9 10 11 12 | test=# explain analyze update t1 set name1=name2 from t2 where id1=id2;
QUERY PLAN
---------------------------------------------------------------------------------------------------------------------------------
Update on t1 (cost=37693.00..98122.00 rows=1000000 width=67) (actual time=8197.309..8197.312 rows=0 loops=1)
-> Hash Join (cost=37693.00..98122.00 rows=1000000 width=67) (actual time=349.817..1633.896 rows=1000000 loops=1)
Hash Cond: (t2.id2 = t1.id1)
-> Seq Scan on t2 (cost=0.00..20310.00 rows=1000000 width=61) (actual time=0.021..191.730 rows=1000000 loops=1)
-> Hash (cost=20310.00..20310.00 rows=1000000 width=10) (actual time=348.798..348.798 rows=1000000 loops=1)
Buckets: 131072 Batches: 16 Memory Usage: 3594kB
-> Seq Scan on t1 (cost=0.00..20310.00 rows=1000000 width=10) (actual time=0.034..153.882 rows=1000000 loops=1)
Planning Time: 0.780 ms
Execution Time: 8197.543 ms
|
三、结论
对于大批量数据update，基于hash join 的update方法效率上要高效很多。
