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
Riak 2.0における全文検索 と整合性の問題
Search
UENISHI Kota
November 19, 2014
Technology
860
2
Share
Riak 2.0における全文検索 と整合性の問題
WebDB Forum 2014 のBashoジャパンの発表スライドです
UENISHI Kota
November 19, 2014
More Decks by UENISHI Kota
See All by UENISHI Kota
Storage Systems in Preferred Networks
kuenishi
0
83
Metadata Management in Distributed File Systems
kuenishi
2
550
Behind The Scenes: Cloud Native Storage System for AI
kuenishi
2
440
Apache Ozone behind Simulation and AI Industries
kuenishi
0
440
Distributed Deep Learning with Chainer and Hadoop
kuenishi
3
1.3k
A Few Ways to Accelerate Deep Learning
kuenishi
0
1.2k
Introducing Retz
kuenishi
5
1.2k
Introducing Retz and how to develop practical frameworks
kuenishi
3
790
Formalization and Proof of Distributed Systems (ja)
kuenishi
10
6.5k
Other Decks in Technology
See All in Technology
組織的なAI活用を阻む 最大のハードルは コンテキストデザインだった
ixbox
6
1.4k
AI時代に新卒採用、はじめました/junior-engineer-never-die
dmnlk
0
230
DevOpsDays Tokyo 2026 見えない開発現場を、見える投資に変える
rojoudotcom
2
150
システムは「動く」だけでは 足りない - 非機能要件・分散システム・トレードオフの基礎
nwiizo
25
7.8k
DevOpsDays2026 Tokyo Cross-border practices to connect "safety" and "DX" in healthcare
hokkai7go
0
110
Babylon.js を使って試した色々な内容 / Various things I tried using Babylon.js / Babylon.js 勉強会 vol.5
you
PRO
0
270
Hello UUID
mimifuwacc
0
130
"まず試す"ためのDatabricks Apps活用法 / Databricks Apps for Early Experiments and Validation
nttcom
1
230
ログ基盤・プラグイン・ダッシュボード、全部整えた。でも最後は人だった。
makikub
5
1.3k
スクラムを支える内部品質の話
iij_pr
0
350
ストライクウィッチーズ2期6話のエイラの行動が許せないのでPjMの観点から何をすべきだったのかを考える
ichimichi
1
320
バックオフィスPJのPjMをコーポレートITが担うとうまくいく3つの理由
yueda256
1
300
Featured
See All Featured
Data-driven link building: lessons from a $708K investment (BrightonSEO talk)
szymonslowik
1
1k
Odyssey Design
rkendrick25
PRO
2
570
The Language of Interfaces
destraynor
162
26k
The State of eCommerce SEO: How to Win in Today's Products SERPs - #SEOweek
aleyda
2
10k
XXLCSS - How to scale CSS and keep your sanity
sugarenia
249
1.3M
How STYLIGHT went responsive
nonsquared
100
6k
Navigating the moral maze — ethical principles for Al-driven product design
skipperchong
2
320
Digital Projects Gone Horribly Wrong (And the UX Pros Who Still Save the Day) - Dean Schuster
uxyall
0
1k
Agile that works and the tools we love
rasmusluckow
331
21k
HU Berlin: Industrial-Strength Natural Language Processing with spaCy and Prodigy
inesmontani
PRO
0
310
Visualizing Your Data: Incorporating Mongo into Loggly Infrastructure
mongodb
49
9.9k
The MySQL Ecosystem @ GitHub 2015
samlambert
251
13k
Transcript
Riak 2.0ʹ͓͚Δશจݕࡧ ͱ߹ੑͷ WebDB Forum 2014 Basho Japan ্ 2014/11/19
ࣗݾհ • ࢄγεςϜྺ6 • Bashoδϟύϯͷํ͔Βདྷ·ͨ͠ • Riak CSͷ։ൃ • ͦͷଞຊͷ͜ͱ
BashoͱRiak •ࢄσʔλϕʔεʁ •RiakΛ͍ͬͯΔʁ •BashoΛ͍ͬͯΔʁ
We are hiring •࣮ੈքͷࢄγεςϜͷ ʹڵຯ͋Δਓʂ •@BashoJapan •
[email protected]
•APࢦͷσʔλϕʔεʢ݁Ռ߹ੑʣ •҆ఆੑɺ༧ଌՄೳੑ •ʮσʔλΛઈରʹͳ͘͞ͳ͍ʯ •৽͍͠ݕࡧΤϯδϯͱͯ͠SolrΛόϯυϧ
͜Μͳͱ͜ΖͰ ಈ͍͍ͯ·͢Riak •Rovio (Angry Birds) •Yahoo!JAPAN ͷΫϥυετϨʔδ •NHS (ΠΪϦε ࠃຽอݥαʔϏε)
•League of Legends (MMORPG) •ۜߦɺήʔϜɺখചɺηϯαʔɺetc…
CAPఆཧͱཧͷDB •ͲΜͳނোʹରͯ͠ (partition tolerance) •σʔλৗʹ߹͓ͯ͠Γ (consistency) •γεςϜ͕ࢭ·Δ͜ͱͳ͍ (availability) ͜ͷ3ͭΛಉ࣌ʹຬͨ͢γεςϜଘࡏ͠ͳ͍
CAP Theorem • C: ෳͷAtomic Objectʹର͢Δ࿈ ଓͨ͠ૢ࡞ (w1, w3, w4,
….) ͕શ ͯಉҰͰ͋Δ͜ͱ (linearizable) • A: Atomic Objectʹૢ࡞ w1, w2, …Λ࣮ߦͰ͖Δ͜ͱ • P: ωοτϫʔΫ͕ΕΔͱAtomic Objectͷϝοηʔδ͕౸ୡ͠ͳ͍
Is the network reliable?
None
None
How Riak Works (mostly about search)
run Riak with Search $ which java $ sudo apt-get
install riak $ echo “search = on” >> /etc/riak/riak.conf $ riak start $ riak-admin bucket-type create tt $ riak-admin bucket-type activate tt $ curl -XPUT http://localhost:8098/search/index/t $ curl -XPUT -H ‘content-type: application/json’ \ http://localhost:8098/types/tt/props \ -d ‘{“props”:{“search_index”:”t”}}’
Consistent Hashing • 160-bit Ωʔۭؒ • ۭؒΛ͢Δ • ύʔςΟγϣϯϊʔ υ͕ݸผཧ
• ϨϓϦΧNݸͷύʔ ςΟγϣϯʹίϐʔ͞ ΕΔ OPEF OPEF OPEF OPEF hash(“meetups/spamham”) N=3
σʔλϕʔεͷݕࡧ • DBࠐܕ • groongaͳͲ • “LIKE” • Pros •
߹ੑΛҡ͍࣋͢͠ • Cons • ີ݁߹ʹͳΔ • ֎෦࿈ܞܕ • Solr • Elasticsearch • Pros • ૄ݁߹ʹͰ͖Δ • Cons • ߹ੑΛҡ࣋͠ʹ͍͘
DBͱΠϯσοΫεͷ߹ੑ • DBͷσʔλຊମͱɺΠϯσοΫεΛ࣌ʹ҆શʹߋ ৽͢ΔͨΊʹτϥϯβΫγϣϯ͕ඞཁ • ີ݁߹ͷγεςϜͰ͋ΕϩοΫͳͲΛͬͯ߹ੑ Λอো͍͢͠
ૄ݁߹ʹ͢ΔϝϦοτ • ಉ͡ݕࡧΤϯδϯͰ͍ΖΜͳDBʹରԠͰ͖Δ • ӡ༻ϊϋπʔϧɺ։ൃΛूதͰ͖Δ • όʔδϣϯ߹ͷ੍͕؇͞ΕΔ • ยํ͕ނো͍ͯͯ͠͏ยํಈ࡞Ͱ͖Δ •
֦ு͘͢͠ͳΔ
྆ํͱ ͍͍ͱ͜ͲΓ͍ͨ͠
Yokozuna
as Riak Search 2.0 = + σʔλͷӬଓԽ Ϋϥελͷཧ γϯϓϧͳૢ࡞ RVFSZ
JOEFYJOH
Riak Search 2.0 • RiakʹSolrΛόϯυϧ • RiakͷKVΛߋ৽͢Δͱσʔλ͕ࣗಈతʹΠϯσο ΫεԽ͞ΕΔ • SolrͷϓϥάΠϯεΩʔϚɺΫΤϦΛͦͷ··
͏͜ͱ͕Ͱ͖Δ • ԽɺϊʔυՃআRiakͷ୲
ૄ݁߹ʁີ݁߹ʁ • ύοέʔδಉࠝ • ϓϩηεͱͯ͠ૄ݁߹ • ΠϯσοΫεͱσʔλؒ݁Ռ߹ੑΛอো • ΠϯσοΫεͷԽ݁Ռ߹ੑΛอো
Solrͱૄ݁߹ • Riak ϓϩηε͕ Solr ͷϓϩηεΛ fork ͢Δ • ཧɺࢹશͯRiak͕ߦ͏
• ผϓϩηεɺϝϞϦۭؒผ
ΠϯσοΫεԽͷྲྀΕ • Riakͷߋ৽ͱҰ৺ಉମʢdocument-based indexingʣ • ΠϯσοΫεԽʹࣦഊ͢ΔͱPUT͕ࣦഊ
ݕࡧΫΤϦͷॲཧ • Riakͷ2iMapReduceͱಉछͷΫΤϦ͕Δ • 3͓͖ͭͷ vnode ʹΞΫηε͢ΔͷͰɺॏෳͳ͍ ू
߹ੑͷ ͱҟৗܥ ߹ੑҡ࣋ͷͨΊʹߋ৽ϩάΛอ࣋ͯ͠ ϦΧόϦ͢Δͱ͍͏ํ๏ͰɺՄ༻ੑͱ அੑΛ୲อͰ͖ͳ͍ɻRiakͰଞͷ ํ๏Ͱσʔλͷ߹ੑΛҡ࣋͢Δɻ
Α͋͘Δҟৗܥ #1 •ϊʔυނোͰෳΛ3ͭ࡞Εͳ͍ •෮چͯ͠ɺ2ͭͷ··ʹͳͬͯ͠·͏
Α͋͘Δҟৗܥ #1 • ෮چͨ͋͠ͱʹRead Repair͢Δ • GETͨ͠ͱ͖ʹɺσʔλʹ͕ܽؕ͋ͬͨΒ RiakଆͰPUT͢͠ v2 v2
not found v2
Read Repair͕ ݺΕͳ͍ͱ σʔλ͕Βͳ͍ʁ YES
ΫϥΠΞϯτ͔Β GET͕ݺΕͳ͍ͱ σʔλ͕Βͳ͍ʁ NO
AAE (Active Anti Entropy) https://www.flickr.com/photos/51pct/7507525118/
Active Anti Entropy • APࢦͷDBͷσʔλྼԽΛ͙ ͨΊͷόοΫάϥϯυॲཧ • Merkle-TreeΛͬͯύʔςΟγϣ ϯຖͷʮνΣοΫαϜʯΛܭࢉ •
ࠩΛݟ͚ͭͨΒͦ͜ΛRead Repair͢Δ hash(vnode=0, pid=0) hash(vnode=1, pid=0) hash(vnode=2, pid=0)
Α͋͘Δҟৗܥ #2 • Solr͕ΠϯσοΫεԽ͢Δલʹམͪͨ • Key, Valueอଘ͞Ε͕ͨΠϯσοΫεͳ͍
Read Repair͕ ݺΕΕΠϯσο Ϋε͕म෮͞ΕΔʁ NO
AAE (Active Anti Entropy) https://www.flickr.com/photos/51pct/7507525118/
YZ Active Anti Entropy • ݕࡧΠϯσοΫεʹ͍ͭͯ Merkle TreeΛ࡞ • ύʔςΟγϣϯຖʹKey-Valueͷ
Treeͱൺֱ͠ɺ͕ࠩ͋ΕΠϯ σοΫεΛमਖ਼ hash(vnode=0, pid=0, kv) hash(vnode=0, pid=0, yz)
Α͋͘Δҟৗܥ #2 •Searchઐ༻AAE͕ಈ࡞͍ͯ͠Δ •ΠϯσοΫεͱσʔλͷෆ߹Λݟ͚ͭͯ म෮͢Δ AAE
Α͋͘Δҟৗܥ #3 • ωοτϫʔΫ͕Ε͍ͯͨͱ͖ʹ྆ଆͰߋ৽͕ى ͖ͨʢSplit Brainʣ • Hinted HandoffʹΑΓͲͪΒॻ͖ࠐΈޭ
Α͋͘Δҟৗܥ #3 • ෮چ࣌ʹHandoffͰฦ͞Εɺ྆ํͷΛอ࣋ • GET·ͨAAEʹΑͬͯRead Repair͞ΕΔ
ҟৗܥ #1, #2, #3 •Ͳͷҟৗɺಛʹۓٸͷରॲඞཁͳ͍ •߹ੑݕࠪͱम෮͕ࣗಈͰߦΘΕΔ •ϧʔνϯϫʔΫͷӡ༻࡞ۀ͕΄ͱΜͲͳ͍
·ͱΊ • σʔλϕʔεͱશจݕࡧͷ • ߹ੑΛͱΔ͔ɺૄ݁߹ΛͱΔ͔ͷ • ৽͍͠Riak Search 2.0 •
ਫฏࢄͯ͠εέʔϧΞτͰ͖Δݕࡧ • ݁Ռ߹ੑΛબͿ͜ͱʹΑΓૄ݁߹ԽͱࢄԽ ʹޭʢ͔͠APܕʹʂʣ • ݁Ռ߹ੑΛબͨ͠ઃܭʹΑΓӡ༻͕؆୯ʹ
Questions?
FAQ •ΦϯϥΠϯͷεΩʔϚมߋʁ •μΠφϛοΫεΩʔϚʁ
outline • σʔλϕʔεʹೖ͍ͬͯΔσʔλΛશจݕࡧ͍ͨ͠→ࢁͷΠϯσοΫε͕Ͱ͖ΔɺFKeyͰ ΠϯσοΫε…ͱ͍͏͚ͩͰ • શจݕࡧΤϯδϯͷछྨ • ֎෦ܕʢ߹ੑΛҡ࣋͠ʹ͍͕͘Ԡ༻͕ޮ͘ɺރΕͨιϑτΣΞϓϥάΠϯ͕ଟ͍ʣ • ΈࠐΈܕʢ߹ੑΛҡ͍͕࣋͢͠ɺϝϯςφϯε͕͍͠ʣ
• Riakͦͷதؒɹόϯυϧͯ͠ΔͷͰָɺ͚ͩͲ࣮ମSolrɺ߹ੑΛࣗಈతʹҡ࣋͢Δ • όϯυϧͷํ๏ɺ߹ੑɺࢄͷํ๏ • ݕࡧΠϯσοΫε͚ͩΛࢄͤ͞Δͷ͍͠ • DBͱΠϯσοΫεͷ߹ɺϨϓϦΧͱϨϓϦΧͷ߹ɺJepsenͷͳ͠