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
クックパッドにおけるAWS GPUインスタンスの利用事例 / Powering by AWS ...
Search
Yuichiro Someya
November 17, 2017
Programming
0
410
クックパッドにおけるAWS GPUインスタンスの利用事例 / Powering by AWS GPU Instances in Cookpad Inc
https://pages.awscloud.com/P3-Launch-Event.html
Yuichiro Someya
November 17, 2017
Tweet
Share
More Decks by Yuichiro Someya
See All by Yuichiro Someya
にんげんがさき 基盤はあと / Developers over ML platform
ayemos
0
14k
機械学習をスモールスタートさせる方法 / small machine learning
ayemos
3
2k
アットホームな分析基盤の作り方 / Homemade Machine Learning Toolkits
ayemos
1
960
サービス開発、機械学習、クラウド / the trinity of machine learning
ayemos
0
3.4k
成長を止めない機械学習のやり方 / Don't stop 'til you get enough (data).
ayemos
15
5.1k
AWS で加速する機械学習 / Accelerate Machine Learning with AWS
ayemos
1
310
クックパッドの機械学習基盤 2018 / Machine Learning Platform at Cookpad ~ 2018 ~
ayemos
15
19k
PyTorchとCaffe2とONNXと深層学習モデルのデプロイについて
ayemos
1
2.9k
How we use GPUs in Cookpad
ayemos
0
140
Other Decks in Programming
See All in Programming
Writing documentation can be fun with plugin system
okuramasafumi
0
120
Software Architecture
hschwentner
6
2.1k
CI改善もDatadogとともに
taumu
0
180
Multi Step Form, Decentralized Autonomous Organization
pumpkiinbell
1
810
How mixi2 Uses TiDB for SNS Scalability and Performance
kanmo
40
16k
Rubyで始める関数型ドメインモデリング
shogo_tksk
0
130
Code smarter, not harder - How AI Coding Tools Boost Your Productivity | Angular Meetup Berlin
danielsogl
0
100
『GO』アプリ データ基盤のログ収集システムコスト削減
mot_techtalk
0
130
第3回関東Kaggler会_AtCoderはKaggleの役に立つ
chettub
3
1.1k
Formの複雑さに立ち向かう
bmthd
1
900
GAEログのコスト削減
mot_techtalk
0
120
新宿駅構内を三人称視点で探索してみる
satoshi7190
2
110
Featured
See All Featured
The Web Performance Landscape in 2024 [PerfNow 2024]
tammyeverts
4
420
Refactoring Trust on Your Teams (GOTO; Chicago 2020)
rmw
33
2.8k
Practical Tips for Bootstrapping Information Extraction Pipelines
honnibal
PRO
12
980
Improving Core Web Vitals using Speculation Rules API
sergeychernyshev
9
500
Put a Button on it: Removing Barriers to Going Fast.
kastner
60
3.7k
How STYLIGHT went responsive
nonsquared
98
5.4k
How GitHub (no longer) Works
holman
314
140k
Let's Do A Bunch of Simple Stuff to Make Websites Faster
chriscoyier
507
140k
実際に使うSQLの書き方 徹底解説 / pgcon21j-tutorial
soudai
175
52k
Testing 201, or: Great Expectations
jmmastey
42
7.2k
Fireside Chat
paigeccino
34
3.2k
How to train your dragon (web standard)
notwaldorf
91
5.8k
Transcript
ΫοΫύουʹ͓͚Δ "84(16Πϯελϯεͷ ར༻ࣄྫ ΫοΫύουגࣜձࣾݚڀ։ൃ෦ છ୩༔Ұ !"NB[PO&$(16ΠϯελϯεࡇΓ
ࣗݾ հ HJUIVCDPNBZFNPT UXJUUFSDPNBZFNPT@Z XXXBZFNPTNF ‣ છ୩༔Ұ<:VJDIJSP4PNFZB> ‣ ౦େܭࢉֶम࢜
‣ ΫοΫύουגࣜձࣾ ݚڀ։ൃ෦ΤϯδχΞ ৽ଔೖࣾ
‣ ΫοΫύουͷ࣮ݧڥʹ͓͚Δ(16ར༻ঢ়گ ‣ 1ΠϯελϯεͷධՁ "HFOEB
Ϩγϐ ສҎ্ ݄࣍ར༻ऀ ສਓ ݄࣌
Χࠃʹల։ ݴޠʹରԠ ݄࣌
7JSUVBM1SJWBUF$MPVE All-in on AWS since 2011 Amazon S3 Amazon
RDS Amazon Redshift
7JSUVBM1SJWBUF$MPVE All-in on AWS since 2011 Amazon S3 Amazon
RDS Amazon Redshift Amazon EC2
ྉཧ͖Ζ͘ ΈࠐΈχϡʔϥϧωοτϫʔΫ ʹΑΔྉཧը૾ͷࣗಈೝࣝ εϚʔτϑΥϯͷࣸਅͷɺ ྉཧࣸਅΛࣗಈతʹऩूه
ྉཧ͖Ζ͘ ΈࠐΈχϡʔϥϧωοτϫʔΫ ʹΑΔྉཧը૾ͷࣗಈೝࣝ εϚʔτϑΥϯͷࣸਅͷɺ ྉཧࣸਅΛࣗಈతʹऩूه IUUQTTQFBLFSEFDLDPNBZFNPTNBDIJOFMFBSOJOHPQTBUDPPLQBE
7JSUVBM1SJWBUF$MPVE All-in on AWS since 2011 Amazon S3 Amazon
RDS Amazon Redshift Amazon EC2
‣ $6%" ‣ DV%// CUDA9 cuDNN7 CUDA8 cuDNN7 CUDA8
cuDNN6
‣ $6%" ‣ DV%// IUUQTBXTBNB[PODPNBNB[POBJBNJT CUDA9 cuDNN7 CUDA8 cuDNN7
CUDA8 cuDNN6
CUDA9 cuDNN7 CUDA8 cuDNN7 CUDA8 cuDNN6 ...
CUDA9 cuDNN7 CUDA8 cuDNN7 CUDA8 cuDNN6 ... ssh "Create"
CUDA9 cuDNN7 CUDA8 cuDNN7 CUDA8 cuDNN6 ... `ssh` ...
CUDA9 cuDNN7 CUDA8 cuDNN7 CUDA8 cuDNN6 ... `ssh` ... API
Gateway Outgoing Webhook {text: "Create instance"} Incoming Webhook "Created! => ssh ayemos-workbench-001" (Lambda Proxy) {text: "Create instance"} 1. Route command 2. Perform operations (on EC2) 3. Send the responce back via webhook
CUDA9 cuDNN7 CUDA8 cuDNN7 CUDA8 cuDNN6 ... `ssh` ... API
Gateway Outgoing Webhook {text: "Create instance"} Incoming Webhook "Created! => ssh ayemos-workbench-001" (Lambda Proxy) {text: "Create instance"} 1. Route command 2. Perform operations (on EC2) 3. Send the responce back via webhook
‣ ΫοΫύουͷ࣮ݧڥʹ͓͚Δ(16ར༻ঢ়گ ‣ 1ΠϯελϯεͷධՁ "HFOEB
‣ ΫοΫύουͷ࣮ݧڥʹ͓͚Δ(16ར༻ঢ়گ ‣ 1ΠϯελϯεͷධՁ "HFOEB 4QPJMFS ʮ1࠷ߴʯ
‣ ࠷ߴ
‣ ΧςΰϦ߹ܭສຕͷը૾Λͬͨ$MBTTJpFSͷֶश (PPH-F/FUΛ༷ ීஈΫοΫύουͰߦΘΕ͍ͯΔΑ͏ͳ࣮ݧʹ͍ۙͷ ‣ FQPDIͷֶशʹֻ͔Δ࣌ؒΛ11Ͱൺֱ
ϕϯνϚʔΫ
1 7 $6%"DV%// 1 , $6%"DV%// #BUDI4J[F
NT N NT NT NT NT NT 00.
‣ ࣮ݧͷ࣮ߦ࣌ؒߴʑ࣌ؒͰ͋Δ͜ͱ͕ଟ͍ .VMUJ(16ͷར༻ɺࢄʹର͢Δχʔζ ݱঢ় ͋·Γߴ͘ͳ͍ ‣ ֶशεΫϦϓτͷ.VMUJ(16ରԠͦΕࣗମ࣮ίετ͕͋Δ ‣ ͳͷͰɺϘʔυ୯ମ͕ૣ͘ͳΔͷͱͯخ͍͠
ߟ
‣ 1ͷ͍͍ͱ͜Ζ &$Ͱ͋Δ ຕຕ͕͍ ·ͱΊ