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
これまでの研究経過と希望する研究 / Research Plan for Doctoral C...
Search
Yuuki Tsubouchi (yuuk1)
February 06, 2020
Research
430
1
Share
Embed
Copy iframe code
Copy JS code
Copy link
Start on current slide
これまでの研究経過と希望する研究 / Research Plan for Doctoral Course
令和2年度4月期入学 京都大学大学院 情報学研究科
博士後期課程 第2次学力検査 口頭試問
Yuuki Tsubouchi (yuuk1)
February 06, 2020
More Decks by Yuuki Tsubouchi (yuuk1)
See All by Yuuki Tsubouchi (yuuk1)
「アラーティング」の話をしよう— SREconや論文等の最先端とのギャップをみる
yuukit
7
2.1k
SAKURAONE:An Open Ethernet-based AI HPC System And Its Observed Workload Dynamicsin a Single-Tenant LLM Development Environment
yuukit
1
720
AIスーパーコンピュータにおけるLLM学習処理性能の計測と可観測性 / AI Supercomputer LLM Benchmarking and Observability
yuukit
1
1.3k
SREはサイバネティクスの夢をみるか? / Do SREs Dream of Cybernetics?
yuukit
3
860
SREのためのテレメトリー技術の探究 / Telemetry for SRE
yuukit
14
4.1k
AIスパコン「さくらONE」の オブザーバビリティ / Observability for AI Supercomputer SAKURAONE
yuukit
2
1.6k
AIスパコン「さくらONE」のLLM学習ベンチマークによる性能評価 / SAKURAONE LLM Training Benchmarking
yuukit
2
1.3k
とあるSREの博士「過程」 / A Certain SRE’s Ph.D. Journey
yuukit
11
7.7k
eBPFを用いたAIネットワーク監視システム論文の実装 / eBPF Japan Meetup #4
yuukit
3
1.9k
Other Decks in Research
See All in Research
最先端NLP勉強会2026 論文紹介:Reasoning with Sampling: Your Base Model is Smarter Than You Think (ICLR 2026 paper)
kogoro
4
940
Santé et fertilité des sols : ce que nous disent les pratiques dans les systèmes de polyculture-élevage
institutdelelevage
PRO
0
150
Développer des solutions de réduction des émissions de méthane entérique : 1ers résultats Méthane 2030
institutdelelevage
PRO
0
280
【令和7年度版】内航船 VS 外航船 早見表
anikisencyo
0
160
Vector Map as Language: Toward Unified Remote Sensing Vector Mapping
satai
3
330
超効率化への挑戦:1bit LLMの現状と展望
yumaichikawa
0
950
Hitting the buffers
beale
0
370
Byzantine Eventual Consistency for Reliable Local-First Software
ept
0
330
Bilan travail en élevage : première restitution des enquêtes 2025-2026
institutdelelevage
PRO
0
260
Adapter les bergeries aux fortes chaleurs
institutdelelevage
PRO
0
170
生成AIなんでも展示会vol6 LT登壇資料 NexteraBERT
rikkabotan7
0
210
SPACE 2026 - Renforcer l’attractivité des métiers des productions animales dans les établissements d'enseignement agricole
institutdelelevage
PRO
0
110
Featured
See All Featured
From π to Pie charts
rasagy
1
430
How GitHub (no longer) Works
holman
316
150k
Leo the Paperboy
mayatellez
10
2.4k
Statistics for Hackers
jakevdp
799
230k
How to audit for AI Accessibility on your Front & Back End
davetheseo
0
620
コードの90%をAIが書く世界で何が待っているのか / What awaits us in a world where 90% of the code is written by AI
rkaga
63
47k
The Cost Of JavaScript in 2023
addyosmani
55
10k
A better future with KSS
kneath
240
18k
Navigating Weather and Climate Data
rabernat
0
570
Agile that works and the tools we love
rasmusluckow
331
22k
Paper Plane
katiecoart
PRO
5
54k
Groundhog Day: Seeking Process in Gaming for Health
codingconduct
0
450
Transcript
͜Ε·Ͱͷݚڀܦաͱر͢Δݚڀ ௶ ༎थ 20202݄6 ྩ24݄ظೖֶ ژେֶେֶӃ ใֶݚڀՊ ത࢜ޙظ՝ఔ ୈֶ̎࣍ྗݕࠪ ޱ಄ࢼ
2 1. ֶ࢜՝ఔɾम࢜՝ఔͰͷݚڀ 2. ر͢Δݚڀͷഎܠͱత 3. ر͢Δݚڀͷ՝ͱํ๏ 4. ظ͞ΕΔݚڀՌ ࣍
1. ֶ࢜՝ఔɾम࢜՝ఔͰͷݚڀ
4 ֶ࢜՝ఔ: SHA-1ܭࢉγεςϜͷߴεϧʔϓοτԽ ɾഎܠ: ετϨʔδ্ͷσʔλอଘྔ͕૿େ͍ͯ͠Δ ɾ: σʔλͷॏෳՕॴΛഉআ͢Δ͜ͱʹΑΓɼσʔλอଘྔΛ ݮՄೳ͕ͩɼॏෳ෦ͷൃݟॲཧ͕ϘτϧωοΫͱͳΔ ɾఏҊ: ෳͷҟͳΔνϟϯΫʹର͢ΔॏෳൃݟॲཧΛSIMDԋࢉثʹ
ΑΓฒྻॲཧ͠ɼߴԽ͢Δ ɾ࣮ݧ: ઌߦख๏ͱൺֱͯ͠ɼ2.0ഒͷεϧʔϓοτ্Λୡͨ͠ ௶༎थ, ҏจ, ஔాਅੜ, ࢁ૱, ദַ, ഡݪ݉Ұ, ॏෳഉআετϨʔδͷͨΊͷ SHA-1ܭࢉγεςϜͷSSE໋ྩʹ ΑΔߴεϧʔϓοτԽ, ిࢠใ௨৴ֶձจࢽ D, 96(10), 2101-2109 2013.
5 म࢜՝ఔ: TCP/IPνΣοΫαϜͷGPUʹΑΔੑೳ্ ɾഎܠ: ίϯϐϡʔλωοτϫʔΫ͕ଳҬԽ͍ͯ͠Δ ɾ: OS෦ͷ௨৴ॲཧͷͨΊͷCPUෛՙ͕ߴ·ͬͯ ͍Δ ɾఏҊ: ௨৴ॲཧͷ͏ͪ௨৴σʔλʹର͢ΔνΣοΫαϜܭࢉΛGPU
ʹҕৡ͢Δ͜ͱʹΑΓɼCPUෛՙΛݮͤ͞Δ ɾ࣮ݧ: CPUʹΑΔνΣοΫαϜΛߦ͏ͱൺֱ͠ɼσʔλసૹε ϧʔϓοτ͕࠷େͰ13%্ ௶༎थ, ୩߶, ୩ޱٛ໌, தതོ, দԬໜొ, TCP/IPελοΫʹ͓͚ΔνΣοΫαϜܭࢉͷ GPU Φϑ ϩʔυʹΑΔ ੑೳ্ख๏, ిࢠใ௨৴ֶձٕज़ݚڀใࠂ, NS, ωοτϫʔΫγεςϜ, 113(244), pp.67-72 2013 10݄.
ཧతʹࢄͨ͠ΞϓϦέʔγϣϯͷ ߴ৴པԽͷͨΊͷΞʔΩςΫνϟʹؔ͢Δݚڀ ത࢜՝ఔͰر͢Δݚڀͷ
2. ر͢Δݚڀͷഎܠͱత
8 WebαʔϏεͷཁٻͷߴԽ ・サービスの信頼性の期待値向上 ・冗⻑性の確保,アクセス増に対するスケーラビリティの向上 ɾ୯ҰͷࣄۀऀʹΑΔෳͷαʔϏεల։ ɾڞ௨෦Λ֤αʔϏεͰڞ༗͢Δ͜ͱͰαʔϏεಉ͕࢜௨৴ ɾ10୯ҐͷظؒͷαʔϏεఏڙ ɾػೳͷՃɼιϑτΣΞͷߋ৽ɼγεςϜߏͷ৽ͳͲΛܧଓ ɾར༻ऀͷ͔ΒΫϥυͷωοτϫʔΫԆΛॖ ɾཧతʹࢄͨ͠ڥʹ͓͚ΔࢄΞϓϦέʔγϣϯઃܭ͕ࠓޙඞ
ཁͱͳΔ => ۭؒతӨڹൣғͷ૿େ => ࣌ؒతӨڹൣғͷ૿େ
9 ཧࢄԽʹؔ͢Δҙࣝ γεςϜͷ෦ঢ়ଶͷѲ ͕͘͠ͳΔ ωοτϫʔΫԆͷ૿Ճ ɾγεςϜʹมߋΛՃ͑Δલ ʹɼϦεΫͷൣғΛݟੵΕ ͳ͘ͳΔ ɾϦεΫ͕ݦࡏԽͨ͠ͱ͖ʹɼ ݪҼͷಛఆ͕͘ͳΔ
ɾ֤ڌʹࢄ͢Δσʔλͷ ಉظ͕ͪ࣌ؒେ͖͘ͳΔ ɾσʔλϕʔε(DBMS)ͷҰ؏ ੑͱੑೳΛཱ྆ͤͮ͞Β͍ ߏཁૉͷ૿Ճ
10 ઌߦݚڀͱ՝: Մ؍ଌੑ ࣌ؒ࣠ํͷՄ؍ଌੑ ۭؒ࣠ํͷՄ؍ଌੑ ɾ࣌ܥྻσʔλϕʔεʹ֤छ ܭଌΛอଘ͢Δ ɾੑೳΛॏࢹ࣮ͯ͠Λ࣌ܥ ྻσʔλʹ࠷దԽ͍ͯ͠Δ ɾطଘͷDBMSΛར༻ͨ͠ޓ
ੑ֦ுੑ͕ࣦΘΕΔ ɾαϒγεςϜؒͷґଘؔΛ ࣗಈͰ͢Δ ɾطଘͷΞϓϦέʔγϣϯίʔ υͷมߋΛͱͳ͏ɼ͋Δ͍ ΞϓϦέʔγϣϯʹ༩͑Δ ੑೳӨڹ͕େ͖͍
11 ઌߦݚڀͱ՝: ੑೳͱσʔλҰ؏ੑ ɾ֤ڌؒͰσʔλΛಉظߋ৽ ɾωοτϫʔΫԆͷେ͖͍ڌ ʹಉظ͕࣌ؒ͞ΕΔ ɾ֤ڌؒͷσʔλΛඇಉظߋ৽ ɾσʔλҰ؏ੑʹର͢ΔΞϓϦ έʔγϣϯ։ൃऀͷൣғ͕ େ͖͘ͳΔ
ऑ͍Ұ؏ੑ(݁Ռ߹ੑ) Ԇͷେ͖ͳڥͰͷੑೳͱσʔλҰ؏ੑͷཱ྆ ڧ͍Ұ؏ੑ
ݚڀͷత ɾత: ཧతʹࢄͨ͠ΞϓϦέʔγϣϯͷ৴པੑ্ ɾ੍: طଘͷΞϓϦέʔγϣϯϛυϧΣΞͷίʔυΛมߋͤͣɼ ੑೳӨڹΛ͋ͨ͑ͳ͍ ɾ3ͭͷݚڀ՝ 1. ࣌ؒ࣠ͷՄ؍ଌੑ: ࣌ܥྻσʔλϕʔεͷੑೳͱޓੑͷཱ྆
2. ۭؒ࣠ͷՄ؍ଌੑ: ґଘؔΛෛՙ͔ͭཏతʹ 3. σʔλͷҰ؏ੑΛอূͭͭ͠ɼԠੑೳΛ࠷େԽ 12
3. ر͢Δݚڀͷ՝ͱํ๏
՝(1) ࣌ؒ࣠ํͷՄ؍ଌੑ ࣌ܥྻσʔλϕʔεʹ͓͚Δੑೳͱ֦ுੑͷཱ྆
15 ఏҊ(1): ՝ͱఏҊ ࣌ܥྻσʔλϕʔεͷ՝ ɾطଘͷDBMSΛར༻ͨ͠ޓ ੑ͕ࣦΘΕΔ ఏҊ ɾ୯ҰͷDBMSΛσʔλߏ୯ҐͰ ղ͠ɼૄ݁߹Խ ɾΠϯϝϞϦͱΦϯσΟεΫͷKVSΛ
֊Խͯ͠ॲཧͱอଘͷޮԽ ɾ֤σʔλߏʹରͯ͠طଘͷDBMS Λར༻Մೳ In-Memory KVS On—Disk KVS Tiering Single DBMS DBMS Index DBMS Log ఏҊΞʔΩςΫνϟ
՝(2) ۭؒ࣠ํʹ͓͚ΔՄ؍ଌੑ ґଘؔΛ͢Δ্Ͱ ཏੑͱΦʔόʔϔουΛཱ྆
17 ఏҊ(2): ιέοτࢹ ґଘؔͷ՝ ɾطଘͷΞϓϦέʔγϣϯ ίʔυͷมߋΛͱͳ͏ɼ ͋Δ͍ΞϓϦέʔγϣϯ ʹ༩͑ΔੑೳӨڹ͕େ͖͍ ఏҊ ɾશαʔό্ͰɼOSΧʔωϧͷTCP
ଓͷऴͰ͋ΔιέοτΛࢹ ɾιέοτࢹϓϩηεΛஔ͢Δͷ ΈͰՄೳ ɾιέοτࢹΞϓϦέʔγϣϯͷ ௨৴ʹׂΓ͜·ͣɼΦʔόϔου Kernel Process TCP Flows . . . User ιέοτࢹ Process
՝(3) ཧࢄڥʹ͓͚Δσʔλϕʔε ͷҰ؏ੑͱੑೳͷཱ྆
19 ఏҊ(3): σʔλͷಉظൣғΛదԠతʹܾఆ σʔλҰ؏ੑͱੑೳͷ՝ ɾ֤ڌؒͰσʔλΛಉظߋ৽ ɾωοτϫʔΫԆͷେ͖ͳڌ ʹಉظॲཧ͕͞ΕΔ ఏҊ ɾಡΈࠐΈॲཧͷΈͷॖʹண͠ɼ தԝͷڌʹॻ͖ࠐΈΛ͚Δ
ɾશڌͰಉظͤͣʹɼதԝͷڌ͔ ΒωοτϫʔΫԆ͕ҰఆΑΓখ ͍͞Ұ෦ͷڌͷΈಉظ ɾΞϓϦέʔγϣϯ୯ҐͷಡΈॻ͖ൺ ʹج͖ͮಉظൣғΛదԠతʹܾఆ Origin தԝͷڌ Replica Τοδͷڌ Replica ಉظൣғ
4. ظ͞ΕΔݚڀՌ
21 ݚڀશମͷظ͞ΕΔߩݙ ɾطଘͷΞϓϦέʔγϣϯίʔυDBMSΛมߋͤͣɼ͔ͭੑೳʹେ ͖ͳӨڹΛ༩͑ͳ͍ͱ͍͏੍ͷͱ 1. ཧࢄΞϓϦέʔγϣϯͷՄ؍ଌੑΛ্Մೳ 2. ؍ଌ݁ՌΛར༻͠ԠੑೳΛ࠷େԽ͢ΔΑ͏ʹదԠతʹ੍ޚՄೳ ɾγεςϜཧऀͷෛ୲Λ૿ͣ͞ʹWebαʔϏεͷ৴པੑΛ্͞ ͤɼਓʑ͕WebαʔϏεΛշద͔ͭ࣋ଓతʹར༻Ͱ͖ΔΑ͏ʹͳΔ