Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Features
Speaker Deck
PRO
Sign in
Sign up for free
Search
Search
Thoughts About Normal and Abnormal Data (PyCon ...
Search
Markus H
October 27, 2017
Technology
0
13k
Thoughts About Normal and Abnormal Data (PyCon UK 2017)
Speaker notes at
https://markusholtermann.eu/2017/10/thoughts-about-normal-and-abnormal-data/
Markus H
October 27, 2017
Tweet
Share
More Decks by Markus H
See All by Markus H
🐍 ❤️ 🦀 — Python loves Rust
markush
0
220
Knock! Knock! Who's There?
markush
0
63
An Introduction To Kubernetes ☸
markush
0
88
Writing Safe Database Migrations (DjangoCon Europe 2021)
markush
0
14k
A Pony On The Move: How Migrations Work In Django 🐎
markush
0
13k
All Hands on Deck — Handling Security Issues
markush
0
13k
Logging Rethought 2: The Actions of Frank Taylor Jr. (PyCon UK 2019)
markush
0
56
Logging Rethought 2: The Actions of Frank Taylor Jr. (PyCon Australia 2019)
markush
1
200
Logging Rethought 2: The Actions of Frank Taylor Jr. (DjangoCon Europe 2019)
markush
0
13k
Other Decks in Technology
See All in Technology
OpenTelemetry Collector internals
ymotongpoo
5
560
Web3 のリアリティ / Web3 Reality
ks91
PRO
0
100
Tenstorrent 開発者プログラム
tenstorrent_japan
0
310
AIにどこまで任せる?実務で使える(かもしれない)AIエージェント設計の考え方
har1101
3
1.2k
Create a Rails8 responsive app with Gemini and RubyLLM
palladius
0
130
(新URLに移行しました)FASTと向き合うことで見えた、大規模アジャイルの難しさと楽しさ
wooootack
0
740
vLLM meetup Tokyo
jpishikawa
1
240
「どこにある?」の解決。生成AI(RAG)で効率化するガバメントクラウド運用
toru_kubota
2
450
「伝える」を加速させるCursor術
naomix
0
640
Claude Code どこまでも/ Claude Code Everywhere
nwiizo
49
29k
AWS全冠したので振りかえってみる
tajimon
0
150
2025/6/21 日本学術会議公開シンポジウム発表資料
keisuke198619
2
410
Featured
See All Featured
How GitHub (no longer) Works
holman
314
140k
The Cost Of JavaScript in 2023
addyosmani
50
8.4k
I Don’t Have Time: Getting Over the Fear to Launch Your Podcast
jcasabona
32
2.3k
Docker and Python
trallard
44
3.4k
A Tale of Four Properties
chriscoyier
159
23k
Design and Strategy: How to Deal with People Who Don’t "Get" Design
morganepeng
130
19k
How to train your dragon (web standard)
notwaldorf
92
6.1k
VelocityConf: Rendering Performance Case Studies
addyosmani
329
24k
Faster Mobile Websites
deanohume
307
31k
The Pragmatic Product Professional
lauravandoore
35
6.7k
Exploring the Power of Turbo Streams & Action Cable | RailsConf2023
kevinliebholz
32
5.9k
Visualization
eitanlees
146
16k
Transcript
Thoughts About Normal and Abnormal Data Markus Holtermann @m_holtermann markusholtermann.eu
@m_holtermann I am Markus Holtermann • Senior Software Engineer at
LaterPay • Django Core Developer
@m_holtermann How do we store our data?
@m_holtermann Files CC-BY-NC 2.0 by Tim Gee https://flic.kr/p/rZm63
@m_holtermann Document Stores CC-BY-SA 4.0 by Susan Gerbic https://commons.wikimedia.org/wiki/File%3AArchive_Room.JPG
@m_holtermann Copyright Geek Batman https://www.youtube.com/watch?v=gPDx_IwdYMY
@m_holtermann Name Home planet Gender Padmé Naboo Female Luke Tatooine
Male Leia Alderaan, Naboo Female
@m_holtermann First Normal Form (1NF)
@m_holtermann PersonID Name Home planet Gender 1 Padmé Naboo Female
2 Luke Tatooine Male 3 Leia Alderaan Female 3 Leia Naboo Female
@m_holtermann PersonID Name Home planet Gender 3 Leia Alderaan Female
3 Leia Naboo Male Update Anomalies
@m_holtermann Second Normal Form (2NF)
@m_holtermann PersonID Name Home planet Gender 1 Padmé Naboo Female
2 Luke Tatooine Male 3 Leia Alderaan Female 3 Leia Naboo Female
@m_holtermann PersonID Planet Name 1 Naboo 2 Tatooine 3 Alderaan
3 Naboo PersonID Name Gender 1 Padmé Female 2 Luke Male 3 Leia Female
@m_holtermann PersonID Planet Name 1 Naboo 2 Tatooine 3 Alderaan
3 Naboo ??? Dagobah Insert Anomalies
@m_holtermann Deletion Anomalies PersonID Planet Name 1 Naboo 2 Tatooine
3 Alderaan 3 Naboo PersonID Name Gender 1 Padmé Female 2 Luke Male 3 Leia Female
@m_holtermann Third Normal Form (3NF)
@m_holtermann PlanetID Name Water 10 Naboo 85% 11 Tatooine 1%
12 Alderaan 78% 13 Dagobah 88% PersonID Name Gender 1 Padmé Female 2 Luke Male 3 Leia Female PersonID PlanetID 1 10 2 11 3 10 3 12
@m_holtermann Database normalization is great!
@m_holtermann Always?
@m_holtermann Yet Another Wiki
@m_holtermann Page + PageID Name Slug Revision + RevisionID PageID
Text Date Database Schema
@m_holtermann Task 1: Fetch a single page and its current
revision
@m_holtermann Task 2: Fetch all page titles and the date
of their current revision
Task 1: Fetch a single page SELECT * FROM page
INNER JOIN revision ON page.page_id = revision.page_id WHERE page.slug = 'some-slug' ORDER BY revision.date DESC LIMIT 1;
Task 2: Fetch all pages SELECT page.name, last_revs.date FROM page
INNER JOIN ( SELECT revision.page_id, MAX(revision.date) date FROM revision GROUP BY revision.page_id ) last_revs ON page.page_id = last_revs.page_id;
@m_holtermann Benchmark Environment • Intel i7-6600U, 2.60GHz • 8 GB
Memory • PostgreSQL 9.6.5 • 10k pages, 6m revisions
@m_holtermann Task 1: Fetch a single page Concurrent queries 10
Pages per connection 1000 Queries per page 10 Queries total 100000
@m_holtermann Task 2: Fetch all pages Concurrent queries 1 Queries
per connection 10 Queries total 10
@m_holtermann Task 1: Fetch a single page
@m_holtermann Task 2: Fetch all pages
@m_holtermann Rae Knowler https://speakerdeck.com/bellisk/unsafe-at-any-speed-pycon-uk-26th-october-2017
@m_holtermann Database Schema Page + PageID Name Slug LastRevision Revision
+ RevisionID PageID Text Date
Task 1: Fetch a single page SELECT * FROM page
INNER JOIN revision ON page.last_revision_id = revision.revision_id WHERE page.slug = 'some-slug';
Task 2: Fetch all pages SELECT page.name, revision.date FROM page
INNER JOIN revision ON page.last_revision_id = revision.revision_id;
@m_holtermann Task 1: Fetch a single page
@m_holtermann Task 2: Fetch all pages
@m_holtermann Conclusion
Thanks Markus Holtermann @m_holtermann markusholtermann.eu