共计 19035 个字符，预计需要花费 48 分钟才能阅读完成。
count(distinct xx) 优化
postgres=# CREATE TABLE t1(f1 serial not null unique,f2 text,f3 text,f4 text,f5 text,f6 text,f7 text,f8 text,f9 text,f10 text,f11 text,f12 text) distribute by shard(f1);
NOTICE: Replica identity is needed for shard table, please add to this table through “alter table” command.
CREATE TABLE
Time: 89.938 ms
postgres=# insert into t1 select t,md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text) from generate_series(1,1000000) as t;
INSERT 0 1000000
Time: 14849.045 ms (00:14.849)
postgres=# analyze t1;
ANALYZE
Time: 1340.387 ms (00:01.340)
postgres=# explain (verbose) select count(distinct f2) from t1;
QUERY PLAN —————————————————————————————————————————————
Aggregate (cost=103320.00..103320.01 rows=1 width=8)
Output: count(DISTINCT f2)
-> Remote Subquery Scan on all (dn01,dn02,dn03,dn04,dn05,dn06,dn07,dn08,dn09,dn10) (cost=100.00..100820.00 rows=1000000 width=33)
Output: f2
-> Seq Scan on public.t1 (cost=0.00..62720.00 rows=1000000 width=33)
Output: f2(6 rows)
Time: 0.748 ms
postgres=# select count(distinct f2) from t1;
count
———
1000000
(1 row)
Time: 6274.684 ms (00:06.275)
postgres=# select count(distinct f2) from t1 where f1 <10;
count
——-
9(1 row)
Time: 19.261 ms
如上 count(distinct f2) 产生在 cn 节点，对于 TP 类业务，须要操作的数据量少的状况下，性能开销没有问题，且比下推执行的性能开销还要小。
但对于一次要操作的数据量比拟大的 AP 类业务，网络传输就会瓶颈，如下是改写后的执行打算：
postgres=# explain (verbose) select count(1) from (select f2 from t1 group by f2) as t ;
QUERY PLAN —————————————————————————————————————————————————————–
Finalize Aggregate (cost=355600.70..355600.71 rows=1 width=8)
Output: count(1)
-> Remote Subquery Scan on all (dn01,dn02,dn03,dn04,dn05,dn06,dn07,dn08,dn09,dn10) (cost=355600.69..355600.70 rows=1 width=0)
Output: PARTIAL count(1)
-> Partial Aggregate (cost=355500.69..355500.70 rows=1 width=8)
Output: PARTIAL count(1)
-> Group (cost=340500.69..345500.69 rows=1000000 width=33)
Output: t1.f2
Group Key: t1.f2
-> Sort (cost=340500.69..343000.69 rows=1000000 width=0)
Output: t1.f2
Sort Key: t1.f2
-> Remote Subquery Scan on all (dn01,dn02,dn03,dn04,dn05,dn06,dn07,dn08,dn09,dn10) (cost=216192.84..226192.84 rows=1000000 width=0)
Output: t1.f2
Distribute results by S: f2
-> Group (cost=216092.84..221092.84 rows=1000000 width=33)
Output: t1.f2
Group Key: t1.f2
-> Sort (cost=216092.84..218592.84 rows=1000000 width=33)
Output: t1.f2
Sort Key: t1.f2
-> Seq Scan on public.t1 (cost=0.00..62720.00 rows=1000000 width=33)
Output: t1.f2(23 rows)
改写后，并行推到 dn 执行，此时查看执行的成果，能够看到对于大量数据计算的 AP 类业务，性能进步了 5 倍。
postgres=# select count(1) from (select f2 from t1 group by f2) as t ;
count
———
1000000
(1 row)
Time: 1328.431 ms (00:01.328)
postgres=# select count(1) from (select f2 from t1 where f1<10 group by f2) as t ;
count
——-
9(1 row)
Time: 24.991 ms
postgres=#
增大 work_mem 缩小 io 拜访
增大 work_mem 后，性能进步了 40 倍，因为 work_mem 足够放下 filter 的数据，不须要再做 Materialize 物化，filter 由原来的 subplan 变成了 hash subplan，间接在内存 hash 表中 filter，性能晋升。
留神，work_mem 默认不宜过大，倡议在某个具体的查问语句中再依据须要进行调整即可。
postgres=# CREATE TABLE t1(f1 serial not null unique,f2 text,f3 text,f4 text,f5 text,f6 text,f7 text,f8 text,f9 text,f10 text,f11 text,f12 text) distribute by shard(f1);
NOTICE: Replica identity is needed for shard table, please add to this table through “alter table” command.
CREATE TABLE
Time: 70.545 ms
postgres=# CREATE TABLE t2(f1 serial not null unique,f2 text,f3 text,f4 text,f5 text,f6 text,f7 text,f8 text,f9 text,f10 text,f11 text,f12 text) distribute by shard(f1);
NOTICE: Replica identity is needed for shard table, please add to this table through “alter table” command.
CREATE TABLE
Time: 61.913 ms
postgres=# insert into t1 select t,md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text) from generate_series(1,1000) as t;
INSERT 0 1000
Time: 48.866 ms
postgres=# insert into t2 select t,md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text) from generate_series(1,50000) as t;
INSERT 0 50000
Time: 792.858 ms
postgres=# analyze t1;
ANALYZE
Time: 175.946 ms
postgres=# analyze t2;
ANALYZE
Time: 318.802 ms
postgres=#
postgres=# explain select * from t1 where f2 not in (select f2 from t2);
QUERY PLAN ———————————————————————————————————————————————–
Remote Subquery Scan on all (dn01,dn02,dn03,dn04,dn05,dn06,dn07,dn08,dn09,dn10) (cost=0.00..2076712.50 rows=500 width=367)
-> Seq Scan on t1 (cost=0.00..2076712.50 rows=500 width=367)
Filter: (NOT (SubPlan 1))
SubPlan 1
-> Materialize (cost=0.00..4028.00 rows=50000 width=33)
-> Remote Subquery Scan on all (dn01,dn02,dn03,dn04,dn05,dn06,dn07,dn08,dn09,dn10) (cost=0.00..3240.00 rows=50000 width=33)
-> Seq Scan on t2 (cost=0.00..3240.00 rows=50000 width=33)(7 rows)
Time: 0.916 ms
postgres=# select * from t1 where f2 not in (select f2 from t2);
| f1 | f2 | f3 | f4 | f5 | f6 | f7 | f8 | f9 | f10 | f11 | f12 |
|---|
(0 rows)
Time: 4226.825 ms (00:04.227)
postgres=# set work_mem to ‘8MB’;
SET
Time: 0.289 ms
postgres=# explain select * from t1 where f2 not in (select f2 from t2);
QUERY PLAN —————————————————————————————————————————————–
Remote Subquery Scan on all (dn01,dn02,dn03,dn04,dn05,dn06,dn07,dn08,dn09,dn10) (cost=3365.00..3577.50 rows=500 width=367)
-> Seq Scan on t1 (cost=3365.00..3577.50 rows=500 width=367)
Filter: (NOT (hashed SubPlan 1))
SubPlan 1
-> Remote Subquery Scan on all (dn01,dn02,dn03,dn04,dn05,dn06,dn07,dn08,dn09,dn10) (cost=0.00..3240.00 rows=50000 width=33)
-> Seq Scan on t2 (cost=0.00..3240.00 rows=50000 width=33)(6 rows)
Time: 0.890 ms
postgres=# select * from t1 where f2 not in (select f2 from t2);
| f1 | f2 | f3 | f4 | f5 | f6 | f7 | f8 | f9 | f10 | f11 | f12 |
|---|
(0 rows)
Time: 105.249 ms
postgres=#
not in 改写为 anti join
上文通过增大计算内存进步性能，但内存不可能有限扩充，如下通过改写语句来进步查问的性能。
postgres=# explain select * from t1 left outer join t2 on t1.f2 = t2.f2 where t2.f2 is null;
QUERY PLAN ————————————————————————————————————————————————-
Remote Subquery Scan on all (dn01,dn02,dn03,dn04,dn05,dn06,dn07,dn08,dn09,dn10) (cost=6405.00..9260.75 rows=1 width=734)
-> Hash Anti Join (cost=6405.00..9260.75 rows=1 width=734)
Hash Cond: (t1.f2 = t2.f2)
-> Remote Subquery Scan on all (dn01,dn02,dn03,dn04,dn05,dn06,dn07,dn08,dn09,dn10) (cost=100.00..682.00 rows=1000 width=367)
Distribute results by S: f2
-> Seq Scan on t1 (cost=0.00..210.00 rows=1000 width=367)
-> Hash (cost=21940.00..21940.00 rows=50000 width=367)
-> Remote Subquery Scan on all (dn01,dn02,dn03,dn04,dn05,dn06,dn07,dn08,dn09,dn10) (cost=100.00..21940.00 rows=50000 width=367)
Distribute results by S: f2
-> Seq Scan on t2 (cost=0.00..3240.00 rows=50000 width=367)(10 rows)
Time: 1.047 ms
postgres=# select * from t1 left outer join t2 on t1.f2 = t2.f2 where t2.f2 is null;
| f1 | f2 | f3 | f4 | f5 | f6 | f7 | f8 | f9 | f10 | f11 | f12 | f1 | f2 | f3 | f4 | f5 | f6 | f7 | f8 | f9 | f10 | f11 | f12 |
|---|
(0 rows)
Time: 107.233 ms
postgres=#
也能够批改 not exists：
postgres=# explain select * from t1 where not exists(select 1 from t2 where t1.f2=t2.f2);
QUERY PLAN ———————————————————————————————————————————————–
Remote Subquery Scan on all (dn01,dn02,dn03,dn04,dn05,dn06,dn07,dn08,dn09,dn10) (cost=3865.00..4078.75 rows=1 width=367)
-> Hash Anti Join (cost=3865.00..4078.75 rows=1 width=367)
Hash Cond: (t1.f2 = t2.f2)
-> Remote Subquery Scan on all (dn01,dn02,dn03,dn04,dn05,dn06,dn07,dn08,dn09,dn10) (cost=100.00..682.00 rows=1000 width=367)
Distribute results by S: f2
-> Seq Scan on t1 (cost=0.00..210.00 rows=1000 width=367)
-> Hash (cost=5240.00..5240.00 rows=50000 width=33)
-> Remote Subquery Scan on all (dn01,dn02,dn03,dn04,dn05,dn06,dn07,dn08,dn09,dn10) (cost=100.00..5240.00 rows=50000 width=33)
Distribute results by S: f2
-> Seq Scan on t2 (cost=0.00..3240.00 rows=50000 width=33)(10 rows)
Time: 0.974 ms
postgres=# select * from t1 where not exists(select 1 from t2 where t1.f2=t2.f2);
| f1 | f2 | f3 | f4 | f5 | f6 | f7 | f8 | f9 | f10 | f11 | f12 |
|---|
(0 rows)
Time: 42.944 ms
postgres=#
散布 key jon+limit 优化
数据筹备：
postgres=# CREATE TABLE t1(f1 serial not null unique,f2 text,f3 text,f4 text,f5 text,f6 text,f7 text,f8 text,f9 text,f10 text,f11 text,f12 text) distribute by shard(f1);
NOTICE: Replica identity is needed for shard table, please add to this table through “alter table” command.
CREATE TABLE
postgres=# CREATE TABLE t2(f1 serial not null unique,f2 text,f3 text,f4 text,f5 text,f6 text,f7 text,f8 text,f9 text,f10 text,f11 text,f12 text) distribute by shard(f1);
NOTICE: Replica identity is needed for shard table, please add to this table through “alter table” command.
CREATE TABLE
postgres=# insert into t1 select t,md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text) from generate_series(1,1000000) as t;
INSERT 0 1000000
postgres=# insert into t2 select t,md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text) from generate_series(1,1000000) as t;
INSERT 0 1000000
postgres=# analyze t1;
ANALYZE
postgres=# analyze t2;
ANALYZE
postgres=#
postgres=# \timing
Timing is on.
postgres=# explain select t1.* from t1,t2 where t1.f1=t2.f1 limit 10;
QUERY PLAN ———————————————————————————————————————————————-
Limit (cost=0.25..1.65 rows=10 width=367)
-> Merge Join (cost=0.25..140446.26 rows=1000000 width=367)
Merge Cond: (t1.f1 = t2.f1)
-> Remote Subquery Scan on all (dn01,dn02,dn03,dn04,dn05,dn06,dn07,dn08,dn09,dn10) (cost=100.12..434823.13 rows=1000000 width=367)
-> Index Scan using t1_f1_key on t1 (cost=0.12..62723.13 rows=1000000 width=367)
-> Remote Subquery Scan on all (dn01,dn02,dn03,dn04,dn05,dn06,dn07,dn08,dn09,dn10) (cost=100.12..71823.13 rows=1000000 width=4)
-> Index Only Scan using t2_f1_key on t2 (cost=0.12..62723.13 rows=1000000 width=4)(7 rows)
Time: 1.372 ms
postgres=# explain analyze select t1.* from t1,t2 where t1.f1=t2.f1 limit 10;
QUERY PLAN ——————————————————————————————————————————————————————————————–
Limit (cost=0.25..1.65 rows=10 width=367) (actual time=2675.437..2948.199 rows=10 loops=1)
-> Merge Join (cost=0.25..140446.26 rows=1000000 width=367) (actual time=2675.431..2675.508 rows=10 loops=1)
Merge Cond: (t1.f1 = t2.f1)
-> Remote Subquery Scan on all (dn01,dn02,dn03,dn04,dn05,dn06,dn07,dn08,dn09,dn10) (cost=100.12..434823.13 rows=1000000 width=367) (actual time=1.661..1.704 rows=10 loops=1)
-> Remote Subquery Scan on all (dn01,dn02,dn03,dn04,dn05,dn06,dn07,dn08,dn09,dn10) (cost=100.12..71823.13 rows=1000000 width=4) (actual time=2673.761..2673.783 rows=10 loops=1)Planning time: 0.358 ms
Execution time: 2973.948 ms
(7 rows)
Time: 2976.008 ms (00:02.976)
postgres=#
以上执行打算是在 cn 上执行，merge join 须要把要 join 的数据拉回 cn 再排序，而后再 join，这里主切的开销在于网络，优化办法是让语句推下去计算，如下所示，两者相差 150 倍的性能，个别状况下，如果须要拉大量的数据回 cn 计算，则下推执行的效率会更好。
postgres=# set prefer_olap to on;
SET
Time: 0.291 ms
postgres=# explain select t1.* from t1,t2 where t1.f1=t2.f1 limit 10;
QUERY PLAN ——————————————————————————————————————————–
Limit (cost=100.25..101.70 rows=10 width=367)
-> Remote Subquery Scan on all (dn01,dn02,dn03,dn04,dn05,dn06,dn07,dn08,dn09,dn10) (cost=100.25..101.70 rows=10 width=367)
-> Limit (cost=0.25..1.65 rows=10 width=367)
-> Merge Join (cost=0.25..140446.26 rows=1000000 width=367)
Merge Cond: (t1.f1 = t2.f1)
-> Index Scan using t1_f1_key on t1 (cost=0.12..62723.13 rows=1000000 width=367)
-> Index Only Scan using t2_f1_key on t2 (cost=0.12..62723.13 rows=1000000 width=4)(7 rows)
Time: 1.061 ms
postgres=# explain analyze select t1.* from t1,t2 where t1.f1=t2.f1 limit 10;
QUERY PLAN —————————————————————————————————————————————————————————
Limit (cost=100.25..101.70 rows=10 width=367) (actual time=1.527..3.899 rows=10 loops=1)
-> Remote Subquery Scan on all (dn01,dn02,dn03,dn04,dn05,dn06,dn07,dn08,dn09,dn10) (cost=100.25..101.70 rows=10 width=367) (actual time=1.525..1.529 rows=10 loops=1)
Planning time: 0.360 ms
Execution time: 18.193 ms
(4 rows)
Time: 19.921 ms
非散布 key join 应用 hash join 性能个别更好
为进步 TP 类业务查问的性能，常常须要对一些字段建设索引，应用有索引字段 join 时，零碎往往也会应用 Merge Cond 和 nestloop。
mydb=# CREATE TABLE t1(f1 serial not null,f2 text,f3 text,f4 text,f5 text,f6 text,f7 text,f8 text,f9 text,f10 text,f11 text,f12 text) distribute by shard(f1);
NOTICE: Replica identity is needed for shard table, please add to this table through “alter table” command.
CREATE TABLE
Time: 481.042 ms
mydb=# create index t1_f1_idx on t1(f2);
CREATE INDEX
Time: 85.521 ms
mydb=# CREATE TABLE t2(f1 serial not null,f2 text,f3 text,f4 text,f5 text,f6 text,f7 text,f8 text,f9 text,f10 text,f11 text,f12 text) distribute by shard(f1);
NOTICE: Replica identity is needed for shard table, please add to this table through “alter table” command.
CREATE TABLE
Time: 75.973 ms
mydb=# create index t2_f1_idx on t2(f2);
CREATE INDEX
Time: 29.890 ms
mydb=# insert into t1 select t,md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text) from generate_series(1,1000000) as t;
INSERT 0 1000000
Time: 16450.623 ms (00:16.451)
mydb=# insert into t2 select t,md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text),md5(t::text) from generate_series(1,1000000) as t;
INSERT 0 1000000
Time: 17218.738 ms (00:17.219)
mydb=# analyze t1;
ANALYZE
Time: 2219.341 ms (00:02.219)
mydb=# analyze t2;
ANALYZE
Time: 1649.506 ms (00:01.650)
mydb=#
–merge join
mydb=# explain select t1.* from t1,t2 where t1.f2=t2.f2 limit 10;
QUERY PLAN ————————————————————————————————————————-
Limit (cost=100.25..102.78 rows=10 width=367)
-> Remote Subquery Scan on all (dn001,dn002) (cost=100.25..102.78 rows=10 width=367)
-> Limit (cost=0.25..2.73 rows=10 width=367)
-> Merge Join (cost=0.25..248056.80 rows=1000000 width=367)
Merge Cond: (t1.f2 = t2.f2)
-> Remote Subquery Scan on all (dn001,dn002) (cost=100.12..487380.85 rows=1000000 width=367)
Distribute results by S: f2
-> Index Scan using t1_f1_idx on t1 (cost=0.12..115280.85 rows=1000000 width=367)
-> Materialize (cost=100.12..155875.95 rows=1000000 width=33)
-> Remote Subquery Scan on all (dn001,dn002) (cost=100.12..153375.95 rows=1000000 width=33)
Distribute results by S: f2
-> Index Only Scan using t2_f1_idx on t2 (cost=0.12..115275.95 rows=1000000 width=33)(12 rows)
Time: 4.183 ms
mydb=# explain analyze select t1.* from t1,t2 where t1.f2=t2.f2 limit 10;
QUERY PLAN ——————————————————————————————————————————————-
Limit (cost=100.25..102.78 rows=10 width=367) (actual time=6555.346..6556.296 rows=10 loops=1)
-> Remote Subquery Scan on all (dn001,dn002) (cost=100.25..102.78 rows=10 width=367) (actual time=6555.343..6555.349 rows=10 loops=1)
Planning time: 0.473 ms
Execution time: 6569.828 ms
(4 rows)
Time: 6614.439 ms (00:06.614)
–nested loop
mydb=# set enable_mergejoin to off;
SET
Time: 0.422 ms
mydb=# explain select t1.* from t1,t2 where t1.f2=t2.f2 limit 10;
QUERY PLAN ——————————————————————————————————————–
Limit (cost=100.12..103.57 rows=10 width=367)
-> Remote Subquery Scan on all (dn001,dn002) (cost=100.12..103.57 rows=10 width=367)
-> Limit (cost=0.12..3.52 rows=10 width=367)
-> Nested Loop (cost=0.12..339232.00 rows=1000000 width=367)
-> Remote Subquery Scan on all (dn001,dn002) (cost=100.00..434740.00 rows=1000000 width=367)
Distribute results by S: f2
-> Seq Scan on t1 (cost=0.00..62640.00 rows=1000000 width=367)
-> Materialize (cost=100.12..100.31 rows=1 width=33)
-> Remote Subquery Scan on all (dn001,dn002) (cost=100.12..100.30 rows=1 width=33)
Distribute results by S: f2
-> Index Only Scan using t2_f1_idx on t2 (cost=0.12..0.27 rows=1 width=33)
Index Cond: (f2 = t1.f2)(12 rows)
Time: 1.033 ms
mydb=# explain analyze select t1.* from t1,t2 where t1.f2=t2.f2 limit 10;
QUERY PLAN ——————————————————————————————————————————————-
Limit (cost=100.12..103.57 rows=10 width=367) (actual time=5608.326..5609.571 rows=10 loops=1)
-> Remote Subquery Scan on all (dn001,dn002) (cost=100.12..103.57 rows=10 width=367) (actual time=5608.323..5608.349 rows=10 loops=1)
Planning time: 0.347 ms
Execution time: 5669.901 ms
(4 rows)
Time: 5672.584 ms (00:05.673)
mydb=# set enable_nestloop to off;
SET
Time: 0.436 ms
mydb=# explain select t1.* from t1,t2 where t1.f2=t2.f2 limit 10;
QUERY PLAN ————————————————————————————————————————-
Limit (cost=85983.00..85984.94 rows=10 width=367)
-> Remote Subquery Scan on all (dn001,dn002) (cost=85983.00..85984.94 rows=10 width=367)
-> Limit (cost=85883.00..85884.89 rows=10 width=367)
-> Hash Join (cost=85883.00..274580.00 rows=1000000 width=367)
Hash Cond: (t1.f2 = t2.f2)
-> Remote Subquery Scan on all (dn001,dn002) (cost=100.00..434740.00 rows=1000000 width=367)
Distribute results by S: f2
-> Seq Scan on t1 (cost=0.00..62640.00 rows=1000000 width=367)
-> Hash (cost=100740.00..100740.00 rows=1000000 width=33)
-> Remote Subquery Scan on all (dn001,dn002) (cost=100.00..100740.00 rows=1000000 width=33)
Distribute results by S: f2
-> Seq Scan on t2 (cost=0.00..62640.00 rows=1000000 width=33)(12 rows)
Time: 1.141 ms
mydb=# explain analyze select t1.* from t1,t2 where t1.f2=t2.f2 limit 10;
QUERY PLAN ———————————————————————————————————————————————–
Limit (cost=85983.00..85984.94 rows=10 width=367) (actual time=1083.691..1085.962 rows=10 loops=1)
-> Remote Subquery Scan on all (dn001,dn002) (cost=85983.00..85984.94 rows=10 width=367) (actual time=1083.688..1083.699 rows=10 loops=1)
Planning time: 0.530 ms
Execution time: 1108.830 ms
(4 rows)
Time: 1117.713 ms (00:01.118)
mydb=#
exists 优化
exists 在数据量比拟大状况下，个别应用的是 Semi Join，在 work_mem 足够大的状况下应用的是 hash join，性能会更好。如下，性能晋升大概一倍。
postgres=# show work_mem;
work_mem
———-
4MB
(1 row)
Time: 0.298 ms
postgres=# explain select count(1) from t1 where exists(select 1 from t2 where t2.t1_f1=t1.f1);
QUERY PLAN ———————————————————————————————————————–
Finalize Aggregate (cost=242218.32..242218.33 rows=1 width=8)
-> Remote Subquery Scan on all (dn001,dn002) (cost=242218.30..242218.32 rows=1 width=0)
-> Partial Aggregate (cost=242118.30..242118.31 rows=1 width=8)
-> Hash Semi Join (cost=110248.00..242118.30 rows=505421 width=0)
Hash Cond: (t1.f1 = t2.t1_f1)
-> Seq Scan on t1 (cost=0.00..17420.00 rows=1000000 width=4)
-> Hash (cost=79340.00..79340.00 rows=3000000 width=4)
-> Remote Subquery Scan on all (dn001,dn002) (cost=100.00..79340.00 rows=3000000 width=4)
Distribute results by S: t1_f1
-> Seq Scan on t2 (cost=0.00..52240.00 rows=3000000 width=4)(10 rows)
Time: 1.091 ms
postgres=# select count(1) from t1 where exists(select 1 from t2 where t2.t1_f1=t1.f1);
count
——–
500000
(1 row)
Time: 3779.401 ms (00:03.779)
postgres=# set work_mem to ‘128MB’;
SET
Time: 0.368 ms
postgres=# explain select count(1) from t1 where exists(select 1 from t2 where t2.t1_f1=t1.f1);
QUERY PLAN ——————————————————————————————————————
Finalize Aggregate (cost=101763.76..101763.77 rows=1 width=8)
-> Remote Subquery Scan on all (dn001,dn002) (cost=101763.75..101763.76 rows=1 width=0)
-> Partial Aggregate (cost=101663.75..101663.76 rows=1 width=8)
-> Hash Join (cost=89660.00..101663.75 rows=505421 width=0)
Hash Cond: (t2.t1_f1 = t1.f1)
-> Remote Subquery Scan on all (dn001,dn002) (cost=59840.00..69443.00 rows=505421 width=4)
Distribute results by S: t1_f1
-> HashAggregate (cost=59740.00..64794.21 rows=505421 width=4)
Group Key: t2.t1_f1
-> Seq Scan on t2 (cost=0.00..52240.00 rows=3000000 width=4)
-> Hash (cost=17420.00..17420.00 rows=1000000 width=4)
-> Seq Scan on t1 (cost=0.00..17420.00 rows=1000000 width=4)(12 rows)
Time: 4.739 ms
postgres=# select count(1) from t1 where exists(select 1 from t2 where t2.t1_f1=t1.f1);
count
——–
500000
(1 row)
Time: 1942.037 ms (00:01.942)
postgres=#