1.9 LAB – Query execution plans (Sakila)

by


1.9 LAB – Query execution plans (Sakila)

This lab illustrates how minor changes in a query may have a significant impact on the execution plan.

MySQL Workbench exercise

Refer to the film, actor, and film_actor tables of the Sakila database. This exercise is based on the initial Sakila installation. If you have altered these tables or their data, your results may be different.

Do the following in MySQL Workbench:

  1. Enter the following statements:
USE sakila;


SELECT last_name, first_name, ROUND(AVG(length), 0) AS average

FROM actor

INNER JOIN film_actor ON film_actor.actor_id = actor.actor_id

INNER JOIN film ON film_actor.film_id = film.film_id

WHERE title = "ALONE TRIP"

GROUP BY last_name, first_name

ORDER BY average;
  1. Highlight the SELECT query.
  2. In the main menu, select Query > Explain Current Statement.
  3. In the Display Info box, highlighted in red below, select Data Read per Join.

Workbench displays the following execution plan:

Image is a screenshot of Workbench. The SELECT query described in the lab instructions is highlighted. Below the SELECT query is a flowchart diagram, representing an execution plan for the SELECT query. The flowchart contains boxes and diamonds labeled 1 through 7. Box 1 has four labels: 1 row, Non-Unique Key Lookup, film, and idx_title. Box 2 has four labels: 5 rows, Non-Unique Key Lookup, film_actor, and idx_fk_film_id. Diamond 3 has two labels: 5 rows, and nested loop. Arrows from boxes 1 and 2 point to diamond 3. Box 4 has four labels: 1 row, Unique Key Lookup, actor, and PRIMARY. Diamond 5 has two labels: 5 rows, and nested loop. Arrows from diamond 3 and box 4 point to diamond 5. Box 6 has two labels: GROUP, and tmp table. Box 7 has two labels: ORDER and filesort. An arrow from box 6 points to box 7. An arrow from box 7 points to an unnumbered box with two labels: Query cost 3.07, and query_block #1.

The execution plan depicts the result of EXPLAIN for the SELECT query. The execution plan has seven steps, corresponding to the red numbers on the screenshot:

  1. Access a single film row using the idx_title index on the title column.
  2. Access matching film_actor rows using the idx_fk_film_id index on the film_id foreign key.
  3. Join the results using the nested loop algorithm.
  4. Access actor rows via the index on the primary key.
  5. Join actor rows with the prior join result using the nested loop algorithm.
  6. Store the result in a temporary table and compute the aggregate function.
  7. Sort and generate the result table.

Refer to MySQL nested loop documentation for an explanation of the nested loop algorithm.

Now, replace = in the WHERE clause with < and generate a new execution plan. Step 1 of the execution plan says Index Range Scan. The index scan accesses all films with titles preceding “ALONE TRIP”, rather than a single film.

Finally, replace < in the WHERE clause with > and generate a third execution plan. Step 1 of the execution plan says Full Table Scan and accesses actor rather than film.

zyLab coding

In the zyLab environment, write EXPLAIN statements for the three queries, in the order described above. Submit the EXPLAIN statements for testing.

The zyLab execution plans do not exactly match the Workbench execution plans, since this lab uses a subset of film, actor, and film_actor rows from the Sakila database.

NOTE: In submit-mode tests that generate multiple result tables, the results are merged. Although the tests run correctly, the results appear in one table.