Understanding query performance

Introduction to Redshift

Jason Myers

Principal Architect

Query optimization process

Check the STL_ALERT_EVENT_LOG table
Run EXPLAIN on your query
Check the SVL_QUERY_SUMMARY and SVL_QUERY_REPORT tables

STL_ALERT_EVENT_LOG table

Contains any Redshift alerts triggered by a query

SELECT * 
  FROM stl_alert_event_log 
 WHERE query = 1337;

 query |             event             |          solution       |     event_time      
=======+===============================+=========================+====================
  1337 | Missing query planner statist | Run the ANALYZE command | 2023-11-13 18:20:58

EXPLAINing the query execution plan

Shows all the steps of a query and their relative costs and rows scanned
Works on all Data Manipulation Language (DML) statements:
- SELECT
- SELECT INTO
- CREATE TABLE AS
- INSERT
- UPDATE
- DELETE

Explain example

-- Running EXPLAIN on our top ten divisions by 
-- revenue query
EXPLAIN WITH top_ten_divisions_by_rev AS
(
  SELECT division_id,
         SUM(revenue) AS revenue_total
    FROM sales_data
   GROUP BY division_id
   ORDER BY revenue_total DESC
   LIMIT 10
),

division_names AS
(
  SELECT id AS division_id,
         name AS division_name
    FROM division_names
)
SELECT division_name,
       revenue_total
  FROM top_ten_divisions_by_rev
  JOIN division_names USING (DIVISION_ID)
 WHERE revenue_total > 100000;

Explain results

|QUERY PLAN                                                                                      |
+------------------------------------------------------------------------------------------------+
|Hash Join  (cost=47.11..58.89 rows=3 width=524)                                                 |
|  Hash Cond: (division_names.id = top_ten_divisions_by_rev.division_id)                         |

|  ->  Seq Scan on division_names  (cost=0.00..11.40 rows=140 width=520)                         |
|  ->  Hash  (cost=47.07..47.07 rows=3 width=12)                                                 |

|        ->  Subquery Scan on top_ten_divisions_by_rev  (cost=46.92..47.07 rows=3 width=12)      |
|              Filter: (top_ten_divisions_by_rev.revenue_total > 100000)                         |
|              ->  Limit  (cost=46.92..46.95 rows=10 width=12)                                   |

|                    ->  Sort  (cost=46.92..47.42 rows=200 width=12)                             |
|                          Sort Key: (sum(sales_data.revenue)) DESC                              |

|                          ->  HashAggregate  (cost=40.60..42.60 rows=200 width=12)              |
|                                Group Key: sales_data.division_id                               |
|                                ->  Seq Scan on sales_data  (cost=0.00..30.40 rows=2040 width=8)|

¹ https://www.postgresql.org/docs/current/using-explain.html

Let's practice!

Introduction to Redshift