Conf orm - explain

The amazing world
behind your ORM
Louise Grandjonc

Louise Grandjonc (louise@ulule.com)
Lead developer at Ulule (www.ulule.com)
Django developer - Postgres enthusiast
@louisemeta on twitter
About me

1. How do we end up with performances problems?
2. How can we see them without roughly guessing how
long you’re waiting before seeing your page?
3. What does it change in our everyday developer job?
Today’s agenda

How do we end up with
performances problems?

1. To annoy the DBAs
2. Because we can avoid having to worry about DB
connections
3. We keep using our main language
4. We are a bit afraid of SQL
5. 90% of the time, we don’t really need to do more than
really simple SELECT and INSERT, so why bother do it worst
than our ORM would?
Why do we use ORMs?
(and why that’s not so terrible)

Not looking at what happens will cause performances problems,
because…
1.The ORMs execute queries that you might not expect
2.Your queries might not be optimised and you won’t know about it
3.To make DBAs to like you, even if you’re using an ORM
Why we should know what our
ORM is doing

How can we see them without roughly
guessing how long you’re waiting
before seeing your page?

How can I see what is
happening when I do stuff?
1. Django debug toolbar (to see queries and their explain in your
django view)
Advantages: can be easily included in your django templates
Problems: Does not allow you to see everything (ajax calls !), if
you’re working on an API, you cannot use it!
2. Django devserver : puts all the logs of your database into your
runserver output
Advantages: you’re not missing the ajax calls
3. Simply look at your database logs
Advantages: you can see everything, you won’t be disturbed if
you ever change project/programming languages/framework/
computer, you can configure how you see your logs
Problems: you don’t know where your logs are?

Where are my logs?
owl_conference=# show log_directory ;
log_directory
---------------
pg_log
(1 row)
owl_conference=# show data_directory ;
data_directory
-------------------------
/usr/local/var/postgres
(1 row)
owl_conference=# show log_filename ;
log_filename
-------------------------
postgresql-%Y-%m-%d.log
(1 row)

Having good looking logs
(and logging everything like a crazy owl)
owl_conference=# SHOW config_file;
config_file
-----------------------------------------
/usr/local/var/postgres/postgresql.conf
(1 row)
In your postgresql.conf
log_filename = 'postgresql-%Y-%m-%d.log'
log_statement = 'all'
logging_collector = on
log_min_duration_statement = 0
log_line_prefix = '%t [%p]: [%l-1] user=%u,db=%d,host=%h,app=%a'

Having good looking logs
user=owly,db=owl_conference,host=127.0.0.1,app=owl
LOG: statement: SELECT "owl"."id", "owl"."name",
"owl"."employer_name", "owl"."favourite_food", "owl"."job_id",
"owl"."fur_color" FROM "owl" WHERE "owl"."job_id" = 1 LIMIT 10
user=owly,db=owl_conference,host=127.0.0.1,app=owl
LOG: duration: 0.297 ms
DATABASES = {
'default': {
'ENGINE': 'django.db.backends.postgresql_psycopg2',
'NAME': 'owl_conference',
'USER': 'owly',
'PASSWORD': 'mouseEating',
'HOST': '127.0.0.1',
'OPTIONS': {'application_name': 'owl'}
}
}
Your logs should look like

Yep ! I’ve seen my logs… But …
Where are this queries executed in my code?
Django will always execute your queries when it needs to use the
object !
Let’s take an example…

Example
Template
def index(request):
owls = Owl.objects.filter(employer_name=‘Ulule’)
context = {‘owls': owls}
return render(request, 'owls/index.html', context)
SELECT "owl"."id", "owl"."name", "owl"."employer_name",
"owl"."favourite_food", "owl"."job_id", "owl"."fur_color" FROM
"owl" WHERE "owl"."employer_name" = 'Ulule'
{% for owl in owls %}
<p> {{ owl.name }} </p>
{% end for %}

Example
View
def index(request):
owl_count = len(owls)
context = {‘owls': owls,‘owl_count’: owl_count}
"owl"."favourite_food", "owl"."job_id", "owl"."fur_color" FROM
"owl" WHERE "owl"."employer_name" = 'Ulule'
{% for owl in owls %}
<p> {{ owl.name }} </p>
{% end for %}

Yep ! I’ve seen my logs… But …
Where are this queries executed in my code?
How to spot where your query is executed?
1. Each model has a table to store data.
Find the model.
2. Where in my view, or in my form am I
using this model to get/ﬁlter objects?
3. Where am I using this objects? Is it in my
view/form? Passed into the context and
used in templates?

What does in change in our everyday
developer job?
(Or how to really do something when you have a problem)

The two most common
problems of any ORM user…
1. I have way too many queries… Why ?
2. One of my query is freakin' slow… Why?

Once upon a time… 1000 times
The danger of loops in your code, and how your templates
are making fun of you…
1. Preload stuff ! The ORM is executing the queries
when it needs the data, if your looping over foreign
key, whithout any preload, it will just query every
time it needs the foreign key… Imagine you have a
loop over 1 million objects. Use prefetch_related and
select_related (see next slide)
2. In an ideal world, no query should ever be executed
from your django html template. Every data should
be in your context, you should never have
« surprise » queries from your templates !

Once upon a time… 1000 times
select_related or prefetch_related?
In django, select_related and prefetch_related will help you lower
your amount of query by preloading the foreign keys or many-to-
many.
1. select_related uses a join (only for foreign keys):
- Advantages: only one request
- Problem: if you are joining big tables, with a lot of columns
and no index, it can be slow… We’ll talk about that next.
2. prefetch_related does a second request on your join table (for
foreign keys and many-to-many
- Advantages: no big join
- Problem: more queries

Example … 1
def index(request):
for owl in owls:
# do stuff
owl.job
def index(request):
owls = Owl.objects
.filter(employer_name=‘Ulule’)
.select_related(‘job’)
for owl in owls:
# do stuff
owl.job

Example … 1
Using select_related
owls = Owl.objects
.select_related(‘job’)
"owl"."favourite_food", "owl"."job_id", "owl"."fur_color",
"job"."id", "job"."name" FROM "owl" LEFT OUTER JOIN "job"
ON ("owl"."job_id" = "job"."id") WHERE
"owl"."employer_name" = 'Ulule'

Example … 1
Using prefetch_related
owls = Owl.objects
.prefetch_related(‘job’)
"owl"."favourite_food", "owl"."job_id", "owl"."fur_color"
FROM "owl" WHERE "owl"."employer_name" = 'Ulule'
SELECT "job"."id", "job"."name" FROM "job" WHERE "job"."id"
IN (2)

One of my query is super slow…
Let’s talk about EXPLAIN !

What is EXPLAIN
Gives you the execution plan chosen by the query
planner that your database will use to execute your SQL
statement
Using ANALYZE will actually execute your query! (Don’t
worry, you can ROLLBACK)
EXPLAIN (ANALYZE) my super query;
BEGIN;
EXPLAIN ANALYZE my super query;
ROLLBACK;

Mmmm… Query planner?
The magical thing that generates execution plans for a
query and calculate what is the cost of each plan.
The best one is used to execute your query (hopefully)

So, what does it took like ?
Let’s imagine a slow query… I’m trying to have all the owls
working at Ulule (super rare job for an owl)
Python version
DB version
Owl.objects.filter(employer_name=‘Ulule’)
"owl"."favourite_food", "owl"."job_id", "owl"."fur_color"
FROM "owl" WHERE "owl"."employer_name" = 'Ulule'

And…
owl_conference=# EXPLAIN ANALYZE
SELECT * FROM owl WHERE
employer_name=‘Ulule'
QUERY PLAN
------------------------------------
Seq Scan on owl (cost=0.00..205.01
rows=1 width=35) (actual
time=1.945..1.946 rows=1 loops=1)
Filter: ((employer_name)::text =
'Ulule'::text)
Rows Removed by Filter: 10000
Planning time: 0.080 ms
Execution time: 1.965 ms
(5 rows)

Let’s go step by step ! .. 1
Costs
(cost=0.00..205.01 rows=1 width=35)
Cost of retrieving
all rows
Number of rows
returned
Cost of retrieving
first row
Average width of a
row (in bytes)
(actual time=1.945..1.946 rows=1 loops=1)
Only if you use analyse, gives you the real times
Number of time your seq scan
(index scan etc.) was executed

Seq Scan
Seq Scan on owl ...
Filter: ((employer_name)::text = 'Ulule'::text)
Scan the entire database and retrieve the rows that
correspond to your where clause
It’s okay for small databases but can be very
expensive… Do you need an index?

Index scan
QUERY PLAN
-------------------------------------------------
Index Scan using employer_name_owl on owl
(cost=0.29..8.30 rows=1 width=35) (actual
time=0.034..0.034 rows=1 loops=1)
Index Cond: ((employer_name)::text =
'Ulule'::text)
(4 rows)
What if there is an index on this column?
The index is visited row by row in order to
retrieve the data corresponding to your clause.

owl_conference=# EXPLAIN SELECT * FROM "owl"
WHERE "owl"."employer_name" = 'post office’;
QUERY PLAN
-------------------------------------------------
Seq Scan on owl (cost=0.00..205.01 rows=7001
width=35)
Filter: ((employer_name)::text = 'post
office'::text)
(2 rows)
With an index and a really common value !
It’s quicker for common values for the db to read
all data, than scan the index.

Bitmap Heap Scan
owl_conference=# EXPLAIN SELECT * FROM owl WHERE
owl.employer_name = ‘Hogwarts’;
QUERY PLAN
-------------------------------------------------
Bitmap Heap Scan on owl (cost=47.78..152.78
rows=2000 width=35)
Recheck Cond: ((employer_name)::text =
'Hogwarts'::text)
-> Bitmap Index Scan on employer_name_owl
(cost=0.00..47.28 rows=2000 width=0)
Index Cond: ((employer_name)::text =
'Hogwarts'::text)
(4 rows)
With an index and a common value (but not
too common)

Let’s go step by step ! ..4
Bitmap Heap Scan…
Index scan : goes through your index tuple-pointer one
at a time and reads the data from the pages. Uses the
index order.
Bitmap Heap Scan: orders the tuple-pointer in physical
memory order and go through it.
Avoids little «physical jumps » between pages

So we have 3 types of scan
1. Sequential scan
2. Index scan
3. Bitmap heap scan
And now let’s join stuff !

And now let’s join stuff…
Nested loops
owl_conference=# EXPLAIN ANALYZE SELECT * FROM owl JOIN job
ON (job.id = owl.job_id) WHERE job.id=1;
QUERY PLAN
-------------------------------------------------------------
Nested Loop (cost=blabla) (actual time=blabla)
-> Seq Scan on job (cost=blabla)
-> Seq Scan on owl (costblabla)
Filter: (job_id = 1)
(9 rows)

Nested loops
Used for little tables, can be slow
This image
does not
match
the previous
query ;)

Hash Join
ON (job.id = owl.job_id) WHERE job.id>1;
QUERY PLAN
-------------------------------------------------------------
Hash Join (cost=1.17..318.70 rows=10001 width=56) (actual
time=0.033..36.021 rows=1000 loops=1)
Hash Cond: (owl.job_id = job.id)
-> Seq Scan on owl (cost=blabla(
-> Hash (cost=blabla)
Buckets: 1024 Batches: 1 Memory Usage: 9kB
-> Seq Scan on job (cost=blabla)
Filter: (id > 1)
(10 rows)

Hash Join
Smaller table in hashed because
it has to ﬁt into memory

Merge Join
ON (job.id = owl.id) WHERE owl.id>1;
QUERY PLAN
-------------------------------------------------------------
Merge Join (cost=blabla)
Merge Cond: (owl.id = job.id)
-> Index Scan using owl_pkey on owl (cost=blabla)
Index Cond: (id > 1)
-> Sort (cost=blabla)
Sort Key: job.id
Sort Method: quicksort Memory: 25kB
-> Seq Scan on job (cost=blaba)
(10 rows)

Merge Join
Used for big tables, an index can be
used to avoid sorting

So we have 3 types of joins
1. Nested loop
2. Hash join
3. Merge join
And a last word about
ORDER BY
(last part, I swear !)

And now let’s order stuff…
owl_conference=# EXPLAIN ANALYZE SELECT * FROM owl ORDER BY
owl.job_id, owl.favourite_food;
QUERY PLAN
-------------------------------------------------------------
Sort (cost=844.47..869.47 rows=10001 width=35) (actual
time=7.252..8.090 rows=10001 loops=1)
Sort Key: job_id, favourite_food
Sort Method: quicksort Memory: 1166kB
-> Seq Scan on owl (cost=0.00..180.01 rows=10001
width=35) (actual time=0.017..1.181 rows=10001 loops=1)
(6 rows)
Everything is sorted into the memory (which is why it can costly)

With an index
owl_conference=# EXPLAIN ANALYZE SELECT * FROM owl ORDER BY
owl.job_id, owl.favourite_food;
QUERY PLAN
-------------------------------------------------------------
Index Scan using owl_job_id_favourite_food on owl
(cost=0.29..544.66 rows=10001 width=35) (actual
time=0.016..2.835 rows=10001 loops=1)
(3 rows)
Simply use index order

ORDER BY LIMIT
owl_conference=# EXPLAIN ANALYZE SELECT name, employer_name
FROM owl ORDER BY name LIMIT 10;
QUERY PLAN
-------------------------------------------------------------
-------------------------------------------------------
Limit (cost…) (actual time…)
-> Sort (cost…) (actual time…)
Sort Key: name
Sort Method: top-N heapsort Memory: 25kB
-> Seq Scan on owl (cost=0.00..180.01 rows=10001
width=16) (actual time=0.032..5.856 rows=10002 loops=1)
(7 rows)
Like with quicksort, all the data has to be sorted… Why is the memory
taken so muck smaller?

Top-N heap sort
- A heap (sort of tree) is used with a bounded size
- For each row
- If the heap isn’t full, tuple added at the right place
- If heap is full and value smaller (for ASC) than current
values
- Tuple inserted at the right place, last value popped
- Else value discarded

Top-N heap sort
Data to order … Iterations 1.. 2.. 3
Iteration 10

Top-N heap sort
Example (if it wasn’t clear…)
Inserting new smaller value,
Potter eliminated (Voldy’s dream)
Heap in the end, after sorting
all stuff

Be careful when you ORDER BY !
1. Sorting with sort key without limit or index can be
heavy
2. You might need an index, only EXPLAIN will tell
you

Conclusion
- Looking at your DB logs, whatever your favourite solution is,
will help you build a website with good performances
- Always know where your queries come from
- Careful about loops ! Use prefetch_related and
select_related to avoid O(n) queries
- If you have a slow query, there is no magical solution, look
into explain to understand what’s going wrong and find a
solution

Thank you for your attention !
Any questions?
Owly design: zimmoriarty (https://www.instagram.com/zimmoriarty/)

To go further - sources
Owly design: zimmoriarty (https://www.instagram.com/zimmoriarty/)
https://momjian.us/main/writings/pgsql/optimizer.pdf
https://use-the-index-luke.com/sql/plans-dexecution/
postgresql/operations
http://tech.novapost.fr/postgresql-application_name-django-
settings.html

Conf orm - explain

More Related Content

What's hot

Viewers also liked

Similar to Conf orm - explain

Recently uploaded

Conf orm - explain