Question

I have PostgreSQL 9.2 and MySQL 5.5 (InnoDB) installed on my laptop.
Both database engines using default installation and populated from the same CSV file.
I have ‘sales_reports’ table with ca. 700K rows.

Scenario 1:

• following query:

  select name, year, region, branch from sales_reports group by name,
  year, region, branch;

• PostgreSQL 9.2: Total query runtime: 42.14 sec, 18064 rows retrieved

• PostgreSQL explain:

  Group  (cost=165091.16..174275.61 rows=73476 width=58) (actual time=35196.959..41896.739 rows=18064 loops=1)
  ->  Sort  (cost=165091.16..166928.05 rows=734756 width=58) (actual time=35196.956..41704.549 rows=734756 loops=1)
      Sort Key: name, year, region, branch
      Sort Method: external merge  Disk: 49920kB
      ->  Seq Scan on sales_reports  (cost=0.00..38249.56 rows=734756 width=58) (actual time=0.048..282.331 rows=734756 loops=1)
  Total runtime: 41906.628 ms
  

• MySQL 5.5 : Total query runtime: 4.4 sec, 18064 rows retrieved
• MySQL explain:

  +----+-------------+---------------+------+---------------+------+---------+------+--------+---------------------------------+
  | id | select_type | table         | type | possible_keys | key  | key_len | ref  | rows   | Extra                           |
  +----+-------------+---------------+------+---------------+------+---------+------+--------+---------------------------------+
  |  1 | SIMPLE      | sales_reports | ALL  | NULL          | NULL | NULL    | NULL | 729433 | Using temporary; Using filesort |
  +----+-------------+---------------+------+---------------+------+---------+------+--------+---------------------------------+
  

• PostgreSQL 10x times slower

Scenario 2:

• following query:

  select name, year, region, branch, sum(sale) as sale from sales_reports group by name, year, region, branch;

• PostgreSQL 9.2: Total query runtime: 42.51 sec, 18064 rows retrieved
• PostgreSQL explain:

  GroupAggregate  (cost=165091.16..176847.26 rows=73476 width=64) (actual time=35160.911..42254.060 rows=18064 loops=1)
  ->  Sort  (cost=165091.16..166928.05 rows=734756 width=64) (actual time=35160.489..41857.986 rows=734756 loops=1)
      Sort Key: name, year, region, branch
      Sort Method: external merge  Disk: 54760kB
      ->  Seq Scan on sales_reports  (cost=0.00..38249.56 rows=734756 width=64) (actual time=0.047..296.347 rows=734756 loops=1)
  Total runtime: 42264.790 ms
  

• MySQL 5.5 : Total query runtime: 8.15 sec, 18064 rows retrieved
• MySQL explain:

  +----+-------------+---------------+------+---------------+------+---------+------+--------+---------------------------------+
  | id | select_type | table         | type | possible_keys | key  | key_len | ref  | rows   | Extra                           |
  +----+-------------+---------------+------+---------------+------+---------+------+--------+---------------------------------+
  |  1 | SIMPLE      | sales_reports | ALL  | NULL          | NULL | NULL    | NULL | 729433 | Using temporary; Using filesort |
  +----+-------------+---------------+------+---------------+------+---------+------+--------+---------------------------------+
  

• PostgreSQL 5x times slower

Scenario 3:

• following query:

  select name, year, region, sum(sale) as sale from sales_reports group by name, year, region;

• PostgreSQL 9.2: Total query runtime: 1 sec, 18064 rows retrieved
• PostgreSQL explain:

  HashAggregate  (cost=45597.12..45655.62 rows=5850 width=37) (actual time=758.396..759.756 rows=4644 loops=1)
  ->  Seq Scan on sales_reports  (cost=0.00..38249.56 rows=734756 width=37) (actual time=0.061..116.541 rows=734756 loops=1)
  Total runtime: 760.133 ms

• MySQL 5.5 : Total query runtime: 5.8 sec, 18064 rows retrieved
• MySQL explain:

  +----+-------------+---------------+------+---------------+------+---------+------+--------+---------------------------------+
  | id | select_type | table         | type | possible_keys | key  | key_len | ref  | rows   | Extra                           |
  +----+-------------+---------------+------+---------------+------+---------+------+--------+---------------------------------+
  |  1 | SIMPLE      | sales_reports | ALL  | NULL          | NULL | NULL    | NULL | 729433 | Using temporary; Using filesort |
  +----+-------------+---------------+------+---------------+------+---------+------+--------+---------------------------------+
  

• PostgreSQL 5x times faster

Any ideas why first two scenarios are so slow on PostgreSQL?

BTW, I created indexes for fields I’m using in the query on PostgreSQL, I didn’t create any indexes on MySQL.

Thanks,

Marek

Answer 1

Default postgresql config is rather conservative. For starters, try increasing shared_buffers to 1GB. (Remember about restarting the server for the change to take effect.)

Also, try increasing work_mem until the GroupAggregate switches to HashAggregate in the explain. You can change this without a restart.

A word of warning: Before messing with the settings in production, please read the friendly manual, there are some gotchas involved.