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
A New Concept of Consistency in Distributed Dat...
Search
UENISHI Kota
November 28, 2013
Technology
10
3.8k
A New Concept of Consistency in Distributed Database and Implementation in Riak
Web+DB forum 技術報告 by Basho
UENISHI Kota
November 28, 2013
Tweet
Share
More Decks by UENISHI Kota
See All by UENISHI Kota
Storage Systems in Preferred Networks
kuenishi
0
39
Metadata Management in Distributed File Systems
kuenishi
2
520
Behind The Scenes: Cloud Native Storage System for AI
kuenishi
2
410
Apache Ozone behind Simulation and AI Industries
kuenishi
0
380
Distributed Deep Learning with Chainer and Hadoop
kuenishi
3
1.2k
A Few Ways to Accelerate Deep Learning
kuenishi
0
1.1k
Introducing Retz
kuenishi
5
1.2k
Introducing Retz and how to develop practical frameworks
kuenishi
3
740
Formalization and Proof of Distributed Systems (ja)
kuenishi
10
6.4k
Other Decks in Technology
See All in Technology
2025年になってもまだMySQLが好き
yoku0825
8
3.5k
Kubernetes における cgroup driver のしくみ: runwasi の bugfix より
z63d
2
120
なぜSaaSがMCPサーバーをサービス提供するのか?
sansantech
PRO
6
2.1k
LLM翻訳ツールの開発と海外のお客様対応等への社内導入事例
gree_tech
PRO
0
460
ヘブンバーンズレッドにおける、世界観を活かしたミニゲーム企画の作り方
gree_tech
PRO
0
460
BPaaSにおける人と協働する前提のAIエージェント-AWS登壇資料
kentarofujii
0
110
AIエージェントの活用に重要な「MCP (Model Context Protocol)」とは何か
masayamoriofficial
0
250
Webアクセシビリティ入門
recruitengineers
PRO
3
1.5k
MCPで変わる Amebaデザインシステム「Spindle」の開発
spindle
PRO
3
2.5k
ここ一年のCCoEとしてのAWSコスト最適化を振り返る / CCoE AWS Cost Optimization devio2025
masahirokawahara
1
1.3k
生成AI時代に必要な価値ある意思決定を育てる「開発プロセス定義」を用いた中期戦略
kakehashi
PRO
1
260
【 LLMエンジニアがヒューマノイド開発に挑んでみた 】 - 第104回 Machine Learning 15minutes! Hybrid
soneo1127
0
260
Featured
See All Featured
GraphQLの誤解/rethinking-graphql
sonatard
71
11k
jQuery: Nuts, Bolts and Bling
dougneiner
64
7.9k
Building Adaptive Systems
keathley
43
2.7k
How to Think Like a Performance Engineer
csswizardry
26
1.9k
Design and Strategy: How to Deal with People Who Don’t "Get" Design
morganepeng
131
19k
Code Reviewing Like a Champion
maltzj
525
40k
Balancing Empowerment & Direction
lara
3
610
Automating Front-end Workflow
addyosmani
1370
200k
Exploring the Power of Turbo Streams & Action Cable | RailsConf2023
kevinliebholz
34
6k
Bootstrapping a Software Product
garrettdimon
PRO
307
110k
VelocityConf: Rendering Performance Case Studies
addyosmani
332
24k
The Illustrated Children's Guide to Kubernetes
chrisshort
48
50k
Transcript
ࢄσʔλϕʔεʹ͓͚Δ ৽͍͠߹ੑϞσϧͱ Riakʹ͓͚Δ࣮ 2013 / 11 / 28 WebDB Forum
Basho ্߁ଠ
ࢄσʔλϕʔεʹ͓͚Δ ݹͯ͘৽͍͠߹ੑϞσϧͱ Riakʹ͓͚Δ࣮ 2013 / 11 / 28 WebDB Forum
Basho ্߁ଠ
BashoͱRiak •ࢄσʔλϕʔεʁ •RiakΛ͍ͬͯΔʁ •BashoΛ͍ͬͯΔʁ
CAPఆཧͱཧͷDB •ͲΜͳނোʹରͯ͠ (partition tolerance) •σʔλৗʹ߹͓ͯ͠Γ (consistency) •γεςϜ͕ࢭ·Δ͜ͱͳ͍ (availability) ͜ͷ3ͭΛಉ࣌ʹຬͨ͢γεςϜଘࡏ͠ͳ͍
•Մ༻ੑ (Availability) ͕ಛͷσʔλ ϕʔε •ӡ༻͍͢͠ɺେ͖ͳσʔλͰೖΔ •҆ఆੑɺ༧ଌՄೳੑ •ʮσʔλΛઈରʹͳ͘͞ͳ͍ʯ
͜Μͳͱ͜ΖͰ ಈ͍͍ͯ·͢Riak •Rovio (Angry Birds) •Yahoo!JAPAN ͷΫϥυετϨʔδ •NHS (ΠΪϦε ࠃຽอݥαʔϏε)
•Bump (=>Google) •ۜߦɺήʔϜɺখചɺηϯαʔɺetc…
How Riak Works
Consistent Hashing • 160-bit Ωʔۭؒ • ۭؒΛ͢Δ • ύʔςΟγϣϯϊʔ υ͕ݸผཧ
• ϨϓϦΧNݸͷύʔ ςΟγϣϯʹίϐʔ͞ ΕΔ OPEF OPEF OPEF OPEF hash(“meetups/spamham”) N=3
Consistency͍͠ •ߋ৽ΛࢭΊΔʢAvailabilityΛԼ͛Δʣ͔ɺߋ৽ͷ্ॻ͖Λ ڐ͢ʢσʔλΛࣦ͏ʣ͔͔͠બࢶ͕ͳ͍ Server2 Server1 Server3 PUT V=42 PUT V=0
V=?
ConsistencyͷΘΓʹ •ͱΓ͋͑ͣෳͷόʔδϣϯͷڞଘΛڐ͢ •Ͳͷόʔδϣϯ͕ਖ਼͍͔͠ɺ͘͠Ϛʔδ͢Δ͔ΛRead࣌ʹܾఆ Server2 Server1 Server3 PUT V=42 PUT V=0
V=0 or 42 V=0 V=0 or 42 V=42
APΛ࣮ݱ •ωοτϫʔΫஅ͕ى͖͍ͯͯͱΓ͋͑ͣॻ͖ࠐΈΛڐ͢ Server2 Server1 Server3 PUT V=42 PUT V=0 Server4
෮چͨ͠Βॻ͖͢ ྆ํ͓࣋ͬͯ͘
γϣοϐϯάΧʔτͷྫ •UnionΛͱΕΑ͍ Server2 Server1 Server3 PUT cart=[a,b,d] PUT cart=[a,b,c] union([a,b,c],
[a,b,d]) => [a,b,c,d] [a,b,c] [a,b,c] or [a,b,d] [a,b,d]
ෳόʔδϣϯΛ ڐ͢͜ͱͷ •ϓϩάϥϛϯά͕͍͠ʢτϥϯβΫγϣϯૉ Β͍͠ʣ •ݱ࣮ੈքγϣοϐϯάΧʔτͱΧϯλʔ͚ͩ Ͱͳ͍ •҆શͳMerge, update͕Ͱ͖ΔσʔλߏΛຖճ ߟ͑ͳ͚ΕͳΒͳ͍ •͍ͬͯΔ͏ͪʹࣅͨΑ͏ͳϥΠϒϥϦ͕͋ͪ͜
ͪͰग़དྷ্͕Δ
ͳ͍ͥ͠ͷ͔ʁ •σʔλͷWriteͱWrite͕ೖΕସΘΓ͏ ΔʹSerializableͲ͜Ζ͔WriteҰ؏ ͨ͠ঢ়ଶʹͰ͖ͳ͍ Server2 Server1 Server3 w1 w2 w1
w2 w2 (w1 lost)
Logical Monoticity •σʔλʹର͢ΔՄͳૢ࡞ͷΈΛڐ͢ʂ Data = update(w2, update(w1, Data0)) = update(w1,
update(w2, Data0)) Data = merge(update(w2, Data0), Data)
͑: CRDT •ʮෳՄೳͳՄσʔλܕʯ •Conflict-Free Replicated Data Types •Commutative Replicated Data
Types •… •(Going to be included in Riak 2.0) ) CRDTͷ࡞ऀLogical Monotinicy ͱ͍͏ݴ༿͍ͬͯͳ͍
CRDT in Riak 2.0 •KVSͷVʹʮܕʯΛ࣋ͨͤͯɺܕʹΑͬͯ UpdateͱMergeͷϩδοΫΛܾΊΔ •Read࣌ʹMerge͕αʔόʔଆͰࣗಈతʹ࣮ ߦ͞ΕΔ •ΞϓϦέʔγϣϯܕΛࢦఆ͢Δ͚ͩͰΑ͘ɺ ෳόʔδϣϯͷϋϯυϦϯά͕ෆཁʹͳΔ
CRDT example •PN-Counter •Set •OR-sets •LWW-register •Graph…
PN-Counter •σϞ
PN-Counter • merge • {a: {1,-1}, b: {1,0}, c: {2,0}}
• {a: {0,0}, b: {2, 0}, c: {0, -2}} • => {a: {1,-1}, b:{2,0}, c:{2,-2}} => 2 • update • a͕ {increment, 3} Λड͚͚Δͱ • {a: {4,-1}, b: {1,0}, c: {2,0}}
OR-Sets • merge • {a:{“foo”:true}, b:{“bar”:false}} • + {a:{“foo”:true}, b:{“foo”:false,
“bar”:false}} • => {a:{“foo”:true}, b:{“foo”:false, “bar”:true}} • => [“bar”] • update • add: {a:{}} => +”foo” => {a:{“foo”:false}} • remove: {a: {“foo”:false}} => {a: {“foo”:true}}
OR-Sets •σϞ
Ϣʔεέʔε •ΫϦοΫͷΧϯτ (G-counter) • riak-server/types/counters/buckets/likes/datatypes/basho.com -d 1 •γϣοϐϯάΧʔτ (OR-sets) •ϩάΠϯϢʔβʔ
(PN-counter) •͜ΕΒͷΈ߹Θͤ (map & LWW-register, boolean) •{ name : “basho.com”, likes: 20000, users: 3000, links: [ “basho.co.jp”, “basho.co.uk” ], cool: true }
Ͱ͖ͳ͍͜ͱ •ʮ0Ҏ্ʯͷPN-counter •ϢχʔΫͳIDൃߦ •ͦͷଞCAS͕ඞཁͳσʔλߏͱૢ࡞
·ͱΊ •RiakՄ༻ੑͷ͋Δࢄσʔλϕʔε •ෳͷόʔδϣϯΛಉ࣌ʹอ࣋͢ΔͷΛ ڐ͢͜ͱͰՄ༻ੑΛ୲อ •ΞϓϦ։ൃͷқ͕՝ •CRDTͱ͍͏ܕͷಋೖʹΑΓ؆୯͔ͭ σʔλͷͳ͘ͳΒͳ͍ΈΛ࡞ͬͨ
Questions? •Riak 2.0 Λָ͠Έʹ͍ͯͩ͘͠͞ •Web: http://basho.co.jp •Twitter: @BashoJapan •Me:
[email protected]
•ML:
[email protected]
Useful links http://hal.upmc.fr/docs/00/55/55/88/PDF/techreport.pdf http://arxiv.org/pdf/1210.3368.pdf https://gist.github.com/russelldb/f92f44bdfb619e089a4d http://gsd.di.uminho.pt/members/cbm/ps/scadt3.pdf http://arxiv.org/abs/1011.5808