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
ScaleShift-jp-2019-summer
Search
ryo nakamaru
July 17, 2019
Research
1
220
ScaleShift-jp-2019-summer
ScaleShift の機能概要、Kubernetes クラスタとの連携について
ryo nakamaru
July 17, 2019
Tweet
Share
More Decks by ryo nakamaru
See All by ryo nakamaru
AWSで楽をするサービスメッシュ入門/appmesh-trial
pottava
1
1.6k
reinforce-2019-recap-lt
pottava
2
4.1k
Firecracker とは何か/what is Firecracker
pottava
12
5.5k
ハイブリッド並列 on Kubernetes/hybrid-parallel-program-on-kubernetes
pottava
1
450
AWS Fargate + Code 兄弟で始める継続的デリバリー / Continuous Delivery with AWS Fargate and Code brothers
pottava
12
3.3k
Singularity と NVIDIA GPU Cloud で作る ハイブリッド機械学習環境の構築 / Building a hybrid environment for Machine Learning with Singularity and NGC
pottava
3
1.5k
明日から始めるちょい足し λ / get-started-with-aws-lambda
pottava
4
2.5k
NGC と Singularity によるハイブリッド機械学習環境 / A hybrid environment for Machine Learning with NGC and Singularity
pottava
0
510
NGC × Singularity での機械学習環境/MachineLearning environment with NGC and Singularity
pottava
2
850
Other Decks in Research
See All in Research
CoRL2025速報
rpc
4
3.9k
Open Gateway 5GC利用への期待と不安
stellarcraft
2
170
Pythonでジオを使い倒そう! 〜それとFOSS4G Hiroshima 2026のご紹介を少し〜
wata909
0
1.2k
ロボット学習における大規模検索技術の展開と応用
denkiwakame
1
190
教師あり学習と強化学習で作る 最強の数学特化LLM
analokmaus
2
840
「リアル×スキマ時間」を活用したUXリサーチ 〜新規事業を前に進めるためのUXリサーチプロセスの設計〜
techtekt
PRO
0
240
令和最新技術で伝統掲示板を再構築: HonoX で作る型安全なスレッドフロート型掲示板 / かろっく@calloc134 - Hono Conference 2025
calloc134
0
500
Tiaccoon: Unified Access Control with Multiple Transports in Container Networks
hiroyaonoe
0
400
SREはサイバネティクスの夢をみるか? / Do SREs Dream of Cybernetics?
yuukit
3
330
Proposal of an Information Delivery Method for Electronic Paper Signage Using Human Mobility as the Communication Medium / ICCE-Asia 2025
yumulab
0
120
大規模言語モデルにおけるData-Centric AIと合成データの活用 / Data-Centric AI and Synthetic Data in Large Language Models
tsurubee
1
480
データサイエンティストの業務変化
datascientistsociety
PRO
0
170
Featured
See All Featured
Building Applications with DynamoDB
mza
96
6.9k
How to audit for AI Accessibility on your Front & Back End
davetheseo
0
140
The Illustrated Children's Guide to Kubernetes
chrisshort
51
51k
The AI Revolution Will Not Be Monopolized: How open-source beats economies of scale, even for LLMs
inesmontani
PRO
3
2.9k
HU Berlin: Industrial-Strength Natural Language Processing with spaCy and Prodigy
inesmontani
PRO
0
140
Building a Modern Day E-commerce SEO Strategy
aleyda
45
8.5k
30 Presentation Tips
portentint
PRO
1
190
Design and Strategy: How to Deal with People Who Don’t "Get" Design
morganepeng
133
19k
The Director’s Chair: Orchestrating AI for Truly Effective Learning
tmiket
1
79
The untapped power of vector embeddings
frankvandijk
1
1.5k
My Coaching Mixtape
mlcsv
0
26
A designer walks into a library…
pauljervisheath
210
24k
Transcript
ScaleShift ΦϯϓϨϛε / Ϋϥυ Ͱ࣮ݱ͢Δػցֶशڥ June, 2019
!2 ScaleShift Docker ϕʔεɺΦʔϓϯιʔεͷ Web ΫϥΠΞϯτ ΞϓϦέʔγϣϯͰ͢ • ϞσϧߏஙϑΣʔζ -
NGC / ࣗࣾϦϙδτϦ͔Βػցֶश Docker ΠϝʔδΛϫϯΫϦοΫͰऔಘ - ͦͷҙͷ Docker ΠϝʔδΛ Jupyter notebook ίϯςφͱͯ͠ىಈ • ϞσϧֶशϑΣʔζ - ߏஙʹར༻ͨ͠ϥΠϒϥϦ͝ͱ Docker ΠϝʔδʹݻΊϦϙδτϦอଘ - ΫϦοΫ͚ͩͰ Kubernetes Ϋϥελ / Rescale େنܭࢉλεΫΛૹ৴
جຊతͳಈ͖ !3 How does it work?
!4 ScaleShift ͷىಈ ϩʔΧϧʹ Web αʔόʔ্ཱ͕͕ͪΓ·͢
!5 ػցֶशιϑτΣΞͷΠϯετʔϧ NGC / ϓϥΠϕʔτϨδετϦ ͔ΒϫϯΫϦοΫͰμϯϩʔυ
!6 Jupyter notebook ͰͷϞσϧߏங Jupyter Ͱϥοϓͨ͠ίϯςφ͕͔ΜͨΜʹىಈ ϙʔτ࡞ۀྖҬ ίϯςφ͝ͱʹ ͞Εͨ ΫϦʔϯͳڥ
ɹ.
!7 େنܭࢉͷͨΊͷϥοϐϯά ґଘϥΠϒϥϦιʔείʔυ܈Λ·ͱΊɺͻͱͭͷΠϝʔδʹݻΊ·͢
!8 ࣾΫϥελ / ΫϥυܭࢉλεΫೖ ೖઌʹԠͯ͡ඞཁͳ API ͕࣮ߦ͞Ε·͢ ར༻ϦιʔεྔΛܾΊ ΫϥελλεΫೖ
Kubernetes ࿈ܞ !9 Integration with a kubernetes cluster
!10 ػցֶश ͱ Kubernetes Web ք۾Λத৺ʹίϯςφΦʔέετϨʔγϣϯͷσϑΝΫτʹͳͬͨ k8sɻ ػցֶशͷจ຺Ͱίϯςφར༻͕ΜʹͳΓɺԠ༻ࣄྫ͕૿͍͑ͯ·͢ɻ - NVIDIA
͕ެࣜʹαϙʔτΛද໌ [ GTC 2018 Keynote, March 27 ] - Mercari ML Ops Night Vol.1 [ גࣜձࣾ ϝϧΧϦ / May 23, 2018 ] ɹhttps://mercari.connpass.com/event/85931/presentation/ - Jupyter ͚ͩͰػցֶशΛ࣮αʔϏεల։Ͱ͖Δج൫ [ גࣜձࣾϦΫϧʔτϥΠϑελΠϧ ] ɹhttps://engineer.recruit-lifestyle.co.jp/techblog/2018-10-04-ml-platform/ - KubernetesʹΑΔػցֶशج൫ͷઓ [ גࣜձࣾ Preferred Networks / Dec 4, 2018 ] ɹhttps://www.slideshare.net/pfi/kubernetes-125013757
!11 ScaleShift + Kubernetes ߏྫ ετϨʔδ ཧϊʔυ ܭࢉϊʔυ ࣾωοτϫʔΫ NGC
DockerHub ϓϥΠϕʔτ ϨδετϦ Kubernetes ݚڀ / ։ൃνʔϜ ScaleShift ೖΓ ϩʔΧϧ
!12 1. ػցֶशιϑτΣΞͷબ ετϨʔδ ཧϊʔυ ܭࢉϊʔυ ࣾωοτϫʔΫ NGC DockerHub ϓϥΠϕʔτ
ϨδετϦ Kubernetes ݚڀ / ։ൃνʔϜ GUI ͔ΒબͿ͚ͩͰ μϯϩʔυ͕࢝·Γ·͢
!13 2. Ϟσϧߏங ετϨʔδ ཧϊʔυ ܭࢉϊʔυ ࣾωοτϫʔΫ NGC DockerHub ϓϥΠϕʔτ
ϨδετϦ Kubernetes ݚڀ / ։ൃνʔϜ ScaleShift ͕ ϊʔτϒοΫΛىಈ͠·͢
!14 3. ࣮ߦڥɾೖྗσʔλͷసૹ ετϨʔδ ཧϊʔυ ܭࢉϊʔυ ࣾωοτϫʔΫ NGC DockerHub ϓϥΠϕʔτ
ϨδετϦ Kubernetes ScaleShift ͕෦తʹ ඞཁͳసૹΛߦ͍·͢ ݚڀ / ։ൃνʔϜ
!15 4. େنܭࢉͷ࣮ߦΛࢦࣔ ετϨʔδ ཧϊʔυ ܭࢉϊʔυ ࣾωοτϫʔΫ NGC DockerHub ϓϥΠϕʔτ
ϨδετϦ Kubernetes ݚڀ / ։ൃνʔϜ Kubernetes ͷ Job ͱͯ͠ ܭࢉ݅Λૹ৴͠·͢
!16 5. େنܭࢉͷ࣮ߦ ετϨʔδ ཧϊʔυ ܭࢉϊʔυ ࣾωοτϫʔΫ NGC DockerHub Kubernetes
ϓϥΠϕʔτ ϨδετϦ ݚڀ / ։ൃνʔϜ
!17 6. ܭࢉ݁Ռͷ֬ೝ ετϨʔδ ཧϊʔυ ܭࢉϊʔυ ࣾωοτϫʔΫ NGC DockerHub ϓϥΠϕʔτ
ϨδετϦ Kubernetes ݚڀ / ։ൃνʔϜ
!18 Kubernetes ઃఆ / λεΫ࣮ߦը໘
ScaleShift ͷઃఆ !19 Configurations
!20 ֎෦࿈ܞ ࿈ܞػೳ ઃఆ NVIDIA GPU CLOUD • NVIDIA ࣾͷཧ͢Δػցֶश
Docker Πϝʔδͷ Ұཡ / ৄࡉใऔಘɺΠϝʔδͷμϯϩʔυ ɹAPI Ωʔ & Ϣʔβઃఆ ϓϥΠϕʔτϨδετϦ • ࣗࣾͰཧ͢Δػցֶश Docker Πϝʔδͷ ҰཡใऔಘɺΠϝʔδͷμϯϩʔυ ɹଓઌ & Ϣʔβઃఆ AWS • ػցֶश Docker Πϝʔδͷμϯϩʔυ • ϩʔΧϧϑΝΠϧγεςϜͱ S3 ؒͷσʔλ࿈ܞ ʢ࣮༧ఆʣ Kubernetes • ࣾΫϥελ / ΫϥυͰͷେنܭࢉ࣮ߦ ɹkubecfg Rescale • Rescale ϓϥοτϑΥʔϜͰͷେنܭࢉ࣮ߦ ɹҬࢦఆ & API Ωʔ
!21 ىಈΦϓγϣϯʢൈਮʣ ઃఆ֓ཁ ॳظ SS_JUPYTER_MINIMUM_PORT ίϯςφͷଓϙʔτಈతׂ։࢝൪߸ ɾ30000 SS_LOG_LEVEL ΞϓϦέʔγϣϯͷϩάग़ྗϨϕϧ ɹwarn
SS_WORKSPACE_HOST_DIR ϗετଆͷ࡞ۀσʔλอଘྖҬ ɹͳ͠ʢࢦఆඞਢʣ SS_NGC_REGISTRY_ENDPOINT NGC ଓઌ ɹhttps://registry.nvidia.com SS_NGC_REGISTRY_USER_NAME NGC Ϣʔβʔ໊ ɹ$oauthtoken SS_RESCALE_SINGULARITY_VERSION Rescale Ͱͷ Singularity ϥϯλΠϜόʔδϣϯ ɹ3.2.0 SS_RESCALE_JOB_WALLTIME Rescale ͰͷλεΫ࣮ߦ࠷େ࣌ؒ ɾ3600 docker-compose.yml ʹઃఆΛهࡌɺىಈͰ͖·͢