Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Sign up for free
Menu
Search
Features
All features
Private URLs
Password Protection
Custom URLS
Scheduled publishing
Remove Branding
Restrict embedding
Deck Collections
Notes
Features
All features
Private URLs
Password Protection
Custom URLS
Scheduled publishing
Remove Branding
Restrict embedding
Deck Collections
Notes
Explore
Featured decks
Featured speakers
Programming
Technology
Storyboards
Explore
Featured decks
Featured speakers
Programming
Technology
Storyboards
Pricing
Search
Sign in
Sign up for free
Ludicrous speed
Search
Sergey Vysotskiy
December 09, 2014
Technology
72
0
Share
Embed
Copy iframe code
Copy JS code
Copy link
Start on current slide
Ludicrous speed
Fantastic stories about performance requirements
Sergey Vysotskiy
December 09, 2014
Transcript
121 slides
fantastic stories about performance requirements Ludicrous Speed
None
None
None
None
None
PANIC!
Modern human experimentations
None
None
Big business impact Big response delay
No silver bullets today
Perception of time What happens on internets? What is “OK”
response time? What is “OK” load? Summary PLAN
Perception of time What happens on internets? What is “OK”
response time? What is “OK” load? Summary PLAN
Perception of time What happens on internets? What is “OK”
response time? What is “OK” load? Summary PLAN
Perception of time What happens on internets? What is “OK”
response time? What is “OK” load? Summary PLAN
Perception of time What happens on internets? What is “OK”
response time? What is “OK” load? Summary PLAN
Perception Of Time
100-200 ms 500-1000 ms 2-5 seconds 7-10 seconds instantaneous reaction
uninterrupted reaction some reaction oof…
100-200 ms 500-1000 ms 2-5 seconds 7-10 seconds instantaneous reaction
uninterrupted reaction some reaction oof…
100-200 ms 500-1000 ms 2-5 seconds 7-10 seconds instantaneous reaction
uninterrupted reaction some reaction oof…
100-200 ms 500-1000 ms 2-5 seconds 7-10 seconds instantaneous reaction
uninterrupted reaction some reaction oof…
100-200 ms 500-1000 ms 2-5 seconds 7-10 seconds instantaneous reaction
uninterrupted reaction some reaction oof… *20 years old data
Modern data?
3… may be 5 seconds max
3… may be 5 seconds max Even faster in future
Weber–Fechner law 1 2
Weber–Fechner law 1 2
Weber–Fechner law 1 2
Weber–Fechner law 1 2
Weber–Fechner law 1 2 just noticeable difference (jnd)
Weber–Fechner law just noticeable difference (jnd) For time lengths of
up to 30s - jnd is somewhere between 7% and 18%
Weber–Fechner law Faster Slower
Weber–Fechner law 20s Faster Slower
Weber–Fechner law 20s 24s 16s -20% +20% Faster Slower
“Slightly” faster
Perception of average between t1 and t2 is: sqrt(t1*t2)
Example
Benchmark against 30s 60s less than 36s less than 42s
less than 24s slow same slightly slower faster
Benchmark against 30s 60s less than 36s less than 42s
less than 24s slow same slightly slower faster
Benchmark against 30s 60s less than 36s less than 42s
less than 24s slow same slightly slower faster
Benchmark against 30s 60s less than 36s less than 42s
less than 24s slow same slightly slower definitely faster
Is it worth to optimise? Target performance Unnoticeable degradation
Is it worth to optimise? Target performance Unnoticeable degradation
Is it worth to optimise? Target performance Unnoticeable degradation
Why so serious?
Why so serious?
What Happens on Internets?
None
None
Response time: DNS request Establish TCP connection Generate request to
server Send request to server Server generates response Server sends response Page rendering on browser
Response time: DNS request Establish TCP connection Generate request to
server Send request to server Server generates response Server sends response Page rendering on browser
Response time: DNS request Establish TCP connection Generate request to
server Send request to server Server generates response Server sends response Page rendering on browser
Response time: DNS request Establish TCP connection Generate request to
server Send request to server Server generates response Server sends response Page rendering on browser
Response time: DNS request Establish TCP connection Generate request to
server Send request to server Server generates response Server sends response Page rendering on browser
Response time: DNS request Establish TCP connection Generate request to
server Send request to server Server generates response Server sends response Page rendering on browser
Response time: DNS request Establish TCP connection Generate request to
server Send request to server Server generates response Server sends response Page rendering in a browser
Response time: DNS request Establish TCP connection Generate request to
server Send request to server Server generates response Server sends response Page rendering in a browser
Fun facts! Min time to pass 1/2 of globe: 67ms
Later ;) Server generates response
Let’s talk a bit about that Page rendering in a
browser
In a complex projects 60% of time could be easily
spent on rendering
There’s space to optimise
There’s space to optimise
There’s space to optimise
Time budget: Load and render critical parts first
What Is “OK” Response?
It depends
In general case response longer than 300ms means nobody cares
In general case response longer than 300ms means nobody cares
Usually it’s more than 1 second on a client
It should be fast
It should be fast Bad requirement
It should be faster than 300ms
It should be faster than 300ms Bad requirement
Long tail
95% < 300ms 99% < 500ms
95% < 300ms 99% < 500ms
95% < 300ms 99% < 500ms OK requirement
95% < 300ms 99% < 500ms OK requirement But still
depends
Remember about 20% difference!
Remember to measure competitors!
Oof… Done with time!
Oof… Done with time! There are 1 more to go
What Is “OK” Load?
What measurement units to use?
What is “average user”?
“Average human” have less than 1 arm
Requests Per Second (RPS) is what server understands
How much RPS we should be able to handle?
EXTREMELY HARD QUESTION
Pulse of the Web
Spotify Average Monday in Stockholm
Spotify Specific Monday in Stockholm
None
None
None
None
Spotify Average Monday in Eastern US
Financial Times weekly traffic changes
Financial Times weekly traffic changes Extremely different for different clients
Financial Times weekly traffic changes Extreme daily features
Thanksgiving in US Events could make a big difference
Bombing in Moscow subway 29 march 2010
Bombing in Moscow subway 29 march 2010 First explosion 7:57
Load growth 8:10
Think like a user Look for daily/weakly/yearly cycles
Think like a user Look for daily/weakly/yearly cycles
Event based service? Prepare for extremely huge spikes
Do closed beta And watch for load profile
It’s growing
Very Short Summary
Find out what is OK response time for you Remember
about long tail Find out possible max load Test your assumptions
Find out what is OK response time for you Remember
about long tail Find out possible max load Test your assumptions
Find out what is OK response time for you Remember
about long tail Find out possible max load Test your assumptions
Find out what is OK response time for you Remember
about long tail Find out possible max load Test your assumptions
APPLAUSE
QUESTIONS?
Magic Numbers Some
4 datacenters More than 5K servers ~50 millions of unique
users More than 100 people working on backend infrastructure and operations
DATACENTERS: Ashburn San Jose Stockholm London
QUESTIONS?
Sergey Vysotskiy
[email protected]
More Decks by Sergey Vysotskiy
See All by Sergey Vysotskiy
Embracing the Imposter
ggame
0
73
Pressing Play at Spotify
ggame
0
100
Testing for capacity
ggame
2
250
Software Industry 101 (2013 season): Красочный мир тестирования
ggame
0
140
Latency
ggame
1
210
Швейцарский нож для Android
ggame
0
290
Анатомия ошибки
ggame
0
320
Other Decks in Technology
See All in Technology
AWS Quick 入門 | 目指せ Quick マスター 01
inariku
PRO
2
340
Context Ontology Accelerator 入門 (業務データの意味を、AI エージェントに渡す)
inariku
PRO
3
670
Claude Securityを使ってみた / claude-security-claude-code-meetup-nagoya
tnir
0
110
Gifu WordPress Meetup #200 に向けて、私たちには何ができるだろうか?/ 2026-11-07 Gifu WordPress Meetup
torounit
0
290
Biohub - Cell Tracking During Development 5th place solution
rist
0
150
Datadog の権限統制 〜Org 統合に向けて調べた注意点 / JDDUG-Meetup_24
visional_engineering_and_design
1
120
JAWS_FESTA2026_オンプレ-AWS間プロキシサーバの移行_性能_障害対応で得た実践ノウハウ_高島一晃.pdf
tkazuaki0820
1
120
Kiro Crew 入門 (手を離しても進む仕事) | 目指せ Kiro マスター 11
inariku
PRO
4
740
Agent Harnessはどう作る?事例から学ぶAI Agentの設計思想
sergicalsix
0
290
Bots in Kiro Crew (One point of contact for ongoing work) | Kiro: Zero to Hero 12
inariku
PRO
1
130
あの日、見つかっていなかったらどうなっていたか
kazzpapa3
2
190
並列度を下げたら速くなった — go build の重複コンパイルを実測で見る
tomoaki25
2
1.2k
Featured
See All Featured
Fight the Zombie Pattern Library - RWD Summit 2016
marcelosomers
234
18k
The Hidden Cost of Media on the Web [PixelPalooza 2025]
tammyeverts
2
580
Winning Ecommerce Organic Search in an AI Era - #searchnstuff2025
aleyda
2
2.2k
Conquering PDFs: document understanding beyond plain text
inesmontani
PRO
4
3.3k
YesSQL, Process and Tooling at Scale
rocio
174
15k
svc-hook: hooking system calls on ARM64 by binary rewriting
retrage
2
670
The Invisible Side of Design
smashingmag
301
52k
AI Search: Implications for SEO and How to Move Forward - #ShenzhenSEOConference
aleyda
1
1.4k
Chasing Engaging Ingredients in Design
codingconduct
0
390
Ten Tips & Tricks for a 🌱 transition
stuffmc
1
280
Art, The Web, and Tiny UX
lynnandtonic
304
22k
Measuring & Analyzing Core Web Vitals
bluesmoon
9
1.1k