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
Learning from performance improvements on Graph...
Search
Fumiaki MATSUSHIMA
September 13, 2023
Technology
1
1k
Learning from performance improvements on GraphQL Ruby
https://www.meetup.com/graphql-tokyo/events/295606838/
Meetup #21 Guest talk and lightning talks
Fumiaki MATSUSHIMA
September 13, 2023
Tweet
Share
More Decks by Fumiaki MATSUSHIMA
See All by Fumiaki MATSUSHIMA
Ruby で作る Ruby (物理)
mtsmfm
1
190
GraphQL Ruby benchmark
mtsmfm
1
750
タイムアウトにご用心 / Timeout might break application state
mtsmfm
6
2.5k
Build REST API with GraphQL Ruby
mtsmfm
0
290
GraphQL Ruby をちょっとだけ速くした / Make graphql-ruby faster a bit
mtsmfm
1
680
Gaming PC on GCP
mtsmfm
0
690
How to introduce GraphQL to an existing React-Redux application
mtsmfm
1
220
Canary release in StudySapuri
mtsmfm
0
3k
Analyze Rails CI
mtsmfm
2
810
Other Decks in Technology
See All in Technology
チームが毎日小さな変化と適応を続けたら1年間でスケール可能なアジャイルチームができた話 / Building a Scalable Agile Team
kakehashi
2
220
なぜfreeeはハブ・アンド・スポーク型の データメッシュアーキテクチャにチャレンジするのか?
shinichiro_joya
2
170
I could be Wrong!! - Learning from Agile Experts
kawaguti
PRO
8
3.3k
あなたの知らないクラフトビールの世界
miura55
0
110
2025年のARグラスの潮流
kotauchisunsun
0
790
WantedlyでのKotlin Multiplatformの導入と課題 / Kotlin Multiplatform Implementation and Challenges at Wantedly
kubode
0
240
GoogleのAIエージェント論 Authors: Julia Wiesinger, Patrick Marlow and Vladimir Vuskovic
customercloud
PRO
0
120
Bring Your Own Container: When Containers Turn the Key to EDR Bypass/byoc-avtokyo2024
tkmru
0
840
カップ麺の待ち時間(3分)でわかるPartyRockアップデート
ryutakondo
0
130
Git scrapingで始める継続的なデータ追跡 / Git Scraping
ohbarye
5
470
ドメイン駆動設計の実践により事業の成長スピードと保守性を両立するショッピングクーポン
lycorptech_jp
PRO
6
710
🌏丸い地球を効率的に平たくする 〜🗺️地図の幾何学とWeb地図技術〜
syotasasaki593876
0
140
Featured
See All Featured
How to Think Like a Performance Engineer
csswizardry
22
1.3k
Reflections from 52 weeks, 52 projects
jeffersonlam
348
20k
Distributed Sagas: A Protocol for Coordinating Microservices
caitiem20
330
21k
Measuring & Analyzing Core Web Vitals
bluesmoon
5
210
Build The Right Thing And Hit Your Dates
maggiecrowley
33
2.5k
Faster Mobile Websites
deanohume
305
30k
Building an army of robots
kneath
302
45k
Making the Leap to Tech Lead
cromwellryan
133
9k
Unsuck your backbone
ammeep
669
57k
Creating an realtime collaboration tool: Agile Flush - .NET Oxford
marcduiker
26
1.9k
The Psychology of Web Performance [Beyond Tellerrand 2023]
tammyeverts
45
2.3k
Optimizing for Happiness
mojombo
376
70k
Transcript
@mtsmfm Learning from performance improvements on GraphQL Ruby
Fumiaki Matsushima GitHub, Twitter @mtsmfm
https://www.meetup.com/ja-JP/GraphQL-Tokyo/
https://autify.com/
https://autify.com/careers
https://www.youtube.com/watch?v=dvip2PXi52Y&t=1875s
None
None
Summary of my last talk - GraphQL Ruby 2.0.13 (the
latest version at that time) is twice slower than 1.5.15
Since 2.0.13, there have been 17 releases! https://rubygems.org/gems/graphql/versions
Benchmark target type Article { field0: String! field1: String! ...
fieldN: String! } type Query { articles: [Article!]! } query { articles { field0, field1, ..., fieldN } }
Benchmark target type Article { field0: String! field1: String! ...
fieldN: String! } type Query { articles: [Article!]! } query { articles { field0, field1, ..., fieldN } } Num of articles = 1000 Num of fields = 100
Run benchmark on GitHub Actions https://github.com/mtsmfm/graphql-ruby-benchmark/pull/4
Run benchmark on GitHub Actions https://github.com/mtsmfm/graphql-ruby-benchmark/pull/4
Run benchmark on GitHub Actions https://github.com/mtsmfm/graphql-ruby-benchmark/pull/4
Benchmark env - GitHub Actions (ubuntu-latest) - Ruby 3.2.2 -
https://github.com/mtsmfm/graphql-ruby-benchmark
None
None
Many performance improvements https://github.com/rmosolgo/graphql-ruby/blob/904e8c595637450 40b6d05c5ecb22a73a320c7b5/CHANGELOG.md
None
None
PR #4399 https://github.com/rmosolgo/graphql-ruby/pull/4399
PR #4399 https://github.com/rmosolgo/graphql-ruby/pull/4399
PR #4399 https://github.com/rmosolgo/graphql-ruby/pull/4399
PR #4399 https://github.com/rmosolgo/graphql-ruby/pull/4399
PR #4399 1. Introduces CurrentState class to store runtime state
2. Removes unnecessary context lookup
PR #4399 1. Introduces CurrentState class to store runtime state
2. Removes unnecessary context lookup
Previous code
Previous code
PR #4399 https://github.com/rmosolgo/graphql-ruby/pull/4399
Referring context takes time?!
Context#[] https://github.com/rmosolgo/graphql-ruby/blob/904e8c59563745040b6d05c5ec b22a73a320c7b5/lib/graphql/query/context.rb#L124-L141
Context#[] https://github.com/rmosolgo/graphql-ruby/blob/904e8c59563745040b6d05c5ec b22a73a320c7b5/lib/graphql/query/context.rb#L124-L141 Hash like Hash Set
Looks fast enough 🤔
Benchmark target type Article { field0: String! field1: String! ...
fieldN: String! } type Query { articles: [Article!]! } query { articles { field0, field1, ..., fieldN } } Num of articles = 1000 Num of fields = 100
Benchmark target type Article { field0: String! field1: String! ...
fieldN: String! } type Query { articles: [Article!]! } query { articles { field0, field1, ..., fieldN } } Num of articles = 1000 Num of fields = 100 = 100,000 fields!
Learning from PR #4399 1. Field related code can be
a hotspot a. Even context lookup
Context#[] https://github.com/rmosolgo/graphql-ruby/blob/904e8c59563745040b6d05c5ec b22a73a320c7b5/lib/graphql/query/context.rb#L124-L141
It changes [key] to public_send(key) https://github.com/rmosolgo/graphql-ruby/pull/4399/files#diff-2c8869b38ea62e6 9d0aa4fc7e571e6dc3b8c9f05fba9c25bc79fc17cfee2a335
require 'bundler/inline' gemfile do source 'https://rubygems.org' gem 'benchmark-ips', require: 'benchmark/ips'
gem 'graphql' end state = GraphQL::Execution::Interpreter::Runtime::CurrentState.new Benchmark.ips do |x| x.report('send') do state.public_send(:current_arguments) end x.report('direct') do state.current_arguments end x.compare! end
public_send is 2 times slower $ ruby foo.rb Warming up
-------------------------------------- send 889.676k i/100ms direct 2.157M i/100ms Calculating ------------------------------------- send 8.757M (± 7.4%) i/s - 43.594M in 5.008124s direct 20.391M (± 7.0%) i/s - 103.546M in 5.105176s Comparison: direct: 20391490.9 i/s send: 8756908.0 i/s - 2.33x slower
2% improvements https://github.com/rmosolgo/graphql-ruby/pull/4631
Conclusion 1. GraphQL Ruby is twice faster than last year!
2. Consider total number of fields a. Especially if you’re going to add some codes which affect all fields 3. Enjoy Talk Driven Development