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
PHPで支える大規模アーキテクチャ
Search
yuuki takezawa
August 05, 2017
Technology
1
6.7k
PHPで支える大規模アーキテクチャ
このスライドはPHPカンファレンス関西2017でおこなったものに、presto追加とシンプルなデモを追加したものです
yuuki takezawa
August 05, 2017
Tweet
Share
More Decks by yuuki takezawa
See All by yuuki takezawa
なぜAI時代に 「イベント」を中心に考えるのか? / Why focus on "events" in the age of AI?
ytake
4
1.5k
PHPでアクターモデルを活用したSagaパターンの実践法 / php-saga-pattern-with-actor-model
ytake
0
1.9k
PHP ステートレス VS ステートフル 状態管理と並行性 / php-stateless-stateful
ytake
0
210
PHPでアクターモデルを理解・体験しよう / Understand and experience the actor model in PHP
ytake
2
690
再考 アクターモデル/ reconsider actor model
ytake
0
1.4k
GoとアクターモデルでES+CQRSを実践! / proto_actor_es_cqrs
ytake
1
560
Phluxorでアクターモデルを 理解・体験しよう / toolkit-for-flexible-actor-models-in-php-phluxor
ytake
1
330
オブジェクトのおしゃべり大失敗 メッセージングアンチパターン集 / messaging anti-pattern collection
ytake
2
1.2k
DRE/SREのプラクティス融合によるクラウドネイティブなデータ基盤作り / dre_sre
ytake
0
900
Other Decks in Technology
See All in Technology
VCC 2025 Write-up
bata_24
0
180
10年の共創が示す、これからの開発者と企業の関係 ~ Crossroad
soracom
PRO
1
610
GC25 Recap+: Advancing Go Garbage Collection with Green Tea
logica0419
1
430
フルカイテン株式会社 エンジニア向け採用資料
fullkaiten
0
9.1k
やる気のない自分との向き合い方/How to Deal with Your Unmotivated Self
sanogemaru
0
150
オープンソースでどこまでできる?フォーマル検証チャレンジ
msyksphinz
0
110
社内お問い合わせBotの仕組みと学び
nish01
1
480
生成AI_その前_に_マルチクラウド時代の信頼できるデータを支えるSnowflakeメタデータ活用術.pdf
cm_mikami
0
120
【Kaigi on Rails 事後勉強会LT】MeはどうしてGirlsに? 私とRubyを繋いだRail(s)
joyfrommasara
0
130
社内報はAIにやらせよう / Let AI handle the company newsletter
saka2jp
7
1.2k
自動テストのコストと向き合ってみた
qa
0
200
空間を設計する力を考える / 20251004 Naoki Takahashi
shift_evolve
PRO
4
430
Featured
See All Featured
"I'm Feeling Lucky" - Building Great Search Experiences for Today's Users (#IAC19)
danielanewman
229
22k
Raft: Consensus for Rubyists
vanstee
139
7.1k
Making the Leap to Tech Lead
cromwellryan
135
9.6k
How GitHub (no longer) Works
holman
315
140k
Intergalactic Javascript Robots from Outer Space
tanoku
273
27k
Creating an realtime collaboration tool: Agile Flush - .NET Oxford
marcduiker
32
2.2k
Principles of Awesome APIs and How to Build Them.
keavy
127
17k
The Myth of the Modular Monolith - Day 2 Keynote - Rails World 2024
eileencodes
26
3.1k
Statistics for Hackers
jakevdp
799
220k
CSS Pre-Processors: Stylus, Less & Sass
bermonpainter
358
30k
Context Engineering - Making Every Token Count
addyosmani
5
210
Chrome DevTools: State of the Union 2024 - Debugging React & Beyond
addyosmani
7
890
Transcript
PHPͰࢧ͑Δ େنΞʔΩςΫνϟ ver1.1 takezawa yuuki <ytake> builderscon tokyo 2017
Notice ͜Ε͔Β͓͢͠Δ༰ɺ ॴҦϏοάσʔλʹରͯ͠ͷ ΞϓϩʔνͰ ҰൠతͳΞϓϦέʔγϣϯʹ ͯ·Γ·ͤΜ
Lambda/Kappa Architecture
ҰൠతͳwebαʔϏε • PHP • MySQL, PostgreSQL, Oracle, SQL Server •
Apache, Nginx etc
͓ʁݕࡧ͕͘ͳ͖ͬͯͨͧ Index͖ͪΜͱ͋ΔΜ͚ͩͲɾ
େ͖͘ͳ͖ͬͯͨwebαʔϏε • ઍສϨίʔυʹରͯ͠ϑϩϯτ͔ΒLIKE۟ • ύϑΥʔϚϯεվળͷͨΊʹશจݕࡧͳͲΛՃ͠ RDBMSͷ͍͠ͱ͜ΖΛิ͏ • PVूܭͷͨΊʹຖඦߦ͕ॻ͖ࠐ·ΕΔϩά • খ͍͞αʔϏεͰʹͳΒͳ͔ͬͨͷ͕ݟ͑࢝
ΊΔ
ύλʔϯ1 • webΞϓϦέʔγϣϯଆͰɺσʔλϕʔεʹૠೖޙɺ ElasticseachͳͲʹૠೖ͢Δύλʔϯ • ΞϓϦέʔγϣϯଆͰίϯτϩʔϧͰ͖Δ͕ɺ ΞϓϦέʔγϣϯͷίʔυ͕ංେԽ
Application Database Elasticsearch
/** * @Transactional * * @param ProductEntity $entity */ public
function register(ProductEntity $entity) { $this->productRepository->insert($entity); $this->elasticRepository->index([$entity]); }
ͦͷ2
ύλʔϯ2 • webΞϓϦέʔγϣϯଆͰɺσʔλϕʔεʹૠೖޙɺ ఆظ࣮ߦ͞ΕΔόονͰૠೖ͢Δύλʔϯ • webΞϓϦέʔγϣϯଆͰσʔλϕʔεʹૠೖͷΈ • batchͰͲ͜·Ͱ࡞͔ͨ֬͠ೝ͠ͳ͕Βɺ ະ࡞ͷͷͷΈ࡞͢Δ •
ͨͩ͠ϦΞϧλΠϜͰͳ͍
Application Database Elasticsearch Batch
ൃੜ͢Δ • େྔͷϨίʔυΛऔಘ͢ΔͱςʔϒϧϩοΫ • ϞϊϦγοΫͳγεςϜͰɺಛఆͷσʔλϕʔεʹूத ͨ͠߹ʹɺҶͮΔࣜʹো • େྔϨίʔυҰׅೖͰϨϓϦέʔγϣϯԆͰো • σΟεΫᷓΕͰো
etc…
ύλʔϯ3 • webΞϓϦέʔγϣϯଆͰɺσʔλϕʔεʹૠೖޙɺ ProducerΛհͯ͠Message Queueૠೖ • webΞϓϦέʔγϣϯଆͰσʔλϕʔεʹૠೖͷΈ • Consumer͕Ԡ͠ɺElasticsearchͷindexΛ࡞ •
Message͕ফࣦ͠ͳ͍ݶΓϦΞϧλΠϜʹ͍ۙ
Application Database Elasticsearch Message Queue Consumer Producer
ൃੜͨ͠ • ಛఆͷαʔϏε͕ඞཁͱ͢ΔΛDefinitionʹೖΕͯ͠ ·͍ɺσʔλෆͰQueue٧·Γ • ConsumerͰϝϞϦϦʔΫ • σʔλϕʔείωΫγϣϯΫϩʔζͤͣʹ connection is
gone
ΑΓେ͖ͳΞϓϦέʔγϣϯ
ࣄۀʹΑΔେ͖ͳΞϓϦέʔγϣϯ • ϢʔβʔͷߦಈΛੳ͍ͨ͠ • ଟ͘ͷϢʔβʔʹར༻͞Ε͍ͯΔݕࡧจࣈΛαδΣετ ʹར༻͍ͨ͠ • Ϣʔβʔͷߦಈʹج͍ͮͨίϯςϯπΛද͍ࣔͤͨ͞ • ࢄͨ͠αʔϏεͷσʔλΛूͯ͠৽͍͠ίϯςϯ
πΛఏڙ͍ͨ͠ • BigData
Big Data + Fast Data
BigDataʹ͏ΞϓϦέʔγϣϯͷ՝ • ͦΕͧΕͷΞϓϦέʔγϣϯͰ࣮ߦ͍ͯͨ͠όονॲ ཧ͕ऴΘΒͳ͍ • Ϩίʔυ͍ظؒͰԯͱେʹͳΓɺ σʔλϕʔεͷindexΑΓI/O͕ݫ͍͠ • ϨϓϦέʔγϣϯԆ୲อ͕͍͠ •
ઍສϢʔβʔͷϦΞϧλΠϜͷੳΛ͢Δʹݫ͍͠ • ਖ਼نԽͨ͠σʔλઈରʹRDBMS • ࢄͨ͠σʔλϕʔεʹͲ͏ཱ͔ͪ͏͔
BigDataͷΞϓϩʔν • σʔλͦͷͷͷू • લ·Ͱʹूܭ͓͚ͯ͠ྑ͍σʔλΛ͋Β͔͡Ί༻ ҙ͢Δ • ϦΞϧλΠϜʹೖྗ͞ΕΔσʔλʹରͯ͠ͷ MessageॲཧͱɺࢄՄೳͳσʔλετϨʔδ •
લड़ͷdatabase, elasticsearchซ༻ͷऔΓΈΛɺ ΑΓେ͖ͳεέʔϧͰߏங͢Δ
ࢄετϨʔδ࠾༻ • MongoDBCouchbaseݕ౼ υϥΠόʔपΓͷෆ҆ఆ͞ͳͲͰݟૹΓ(ݱࡏར༻த) • Hadoop ࢄϑΝΠϧγεςϜͷHDFS ࢄॲཧͷͨΊͷMapReduce ेރΕ͍ͯΔɺ࠾༻ࣄྫे
None
ϥϜμΞʔΩςΫνϟ • όονɺαʔϏεɺεϐʔυͰߏ • όονɺେ͖ͳσʔλͷूܭɺେྔσʔλͷੳͳͲΛ୲ ͢Δ -> Hadoop(MapReduce), Spark •
αʔϏεόονͷू݁ՌΛఏڙ͢Δ Hive, HBase, ElephantDB, Splout SQL, pipelineDB… • εϐʔυϦΞϧλΠϜॲཧͷ݁ՌΛఏڙ͢Δ Spark, Storm, Kafka, Cassandra etc.. • αʔϏεͱεϐʔυͷ྆ํͷΛϚʔδͯ͠ฦ٫ қߴ͍ɾɾɾ -> KafkaͳͲʹूͤͨ͞Kappa Architecure
ετϦʔϜॲཧ • େྔͷσʔλΛϦΞϧλΠϜͰॲཧ͢Δͷ͕ɺ ετϦʔϜσʔλॲཧͷత • ऴΘΓ͕ͳ͘ɺແݶʹͬͯ͘ΔͷͷΞϓϩʔν • ϝϞϦͰॲཧ͞Εɺͦͷޙഁغ͞ΕΔ • ࢹܥͷॲཧΑ͘ར༻͞Ε͍ͯΔͷ
• ηϯαʔΛར༻ͨ͠ΞϓϦέʔγϣϯͳͲ
Spark • ࢄॲཧϑϨʔϜϫʔΫͷҰͭ • RDDͱݺΕΔΠϛϡʔλϒϧͳίϨΫγϣϯΛѻ͏ • Spark SQL • Spark
Streaming • Spark MLlib
KappaΞʔΩςΫνϟ
KappaΞʔΩςΫνϟ
OSSͰߏங
PHPϝΠϯͰ࡞Δ͜ͱ͍͠ɾɾ
σϞ https://github.com/ytake/ builderscon-example
Kappa Architecture(small) PHP ConsoleApp Kafka Spark Streaming PHP Consumer Cassandra
PHP WebApp
Apache Cassandra
Apache Cassandra • Ϩίʔυ͕େྔʹ૿͑Δ͜ͱ͕Θ͔͓ͬͯΓɺ ੳʹར༻͢Δ༧ఆͰ͋ͬͨͨΊɺ εέʔϧ͕༰қͱ͍͏Ͱ࠾༻ • PHP͔Βར༻Մೳ(ext-cassandra) • େྔσʔλͷॻ͖ࠐΈʹରԠ
• ؆୯ͳτϥϯβΫγϣϯαϙʔτ • σʔληϯλʔލ͗ͷΫϥελʔߏங • Availability ͱ Partition Tolerance • SQLΠϯλʔϑΣʔε • ୯Ұোͳ͠???????
Apache Cassandra Architecture
ؾΛ͚ͭΔ • RDBMSײ֮Ͱ͏·͘ར༻Ͱ͖·ͤΜ • ύʔςΟγϣϯΩʔͰ͏·͘ઃܭ͢Δ • ݅ʹΑΔΦʔμʔࢦఆͰ͖ͳ͍ • ϚςϦΞϥΠζυϏϡʔซ༻͢͠ •
ো࣌ͷϩάੳͨͩ͘͠ • ίϯύΫγϣϯͱઓ͏(࣮ࡍʹར༻͢Δ༰ྔ*2Ͱܭࢉ) • ݕࡧͰҾ͔͔ͬΔهࣄେମچόʔδϣϯͰɺ ݱߦͱશ͘ผ
ςʔϒϧઃܭ • Primary KeyࣝผΩʔͰ͋Γͳ͕Βɺ Ͳͷnodeʹ֨ೲ͢Δ͔Λܾఆ͢ΔύʔςΟγϣϯΩʔ • ҟͳΔnodeʹ͋Δͷͷݕࡧ͔ͳ͍ ඞཁͳέʔε͕ੜͨ͡߹ςʔϒϧઃܭΛݟ͢ • ߋ৽࣌ɺআ࣌ʹؚΊͳ͚ΕͳΒͳ͍
• ར༻ՄೳͳͷηΧϯμϦΠϯσοΫε·Ͱ • JOINLIKEଘࡏ͠ͳ͍ͨΊɺෳࡶͳͷSparkͰ
ςʔϒϧઃܭ CREATE TABLE timeline.user_timeline ( uuid uuid, user_id int, reference
map<text, text>, body text, is_read tinyint, published_at timestamp, PRIMARY KEY (user_id) );
Ϩίʔυͷॱ൪Λܾఆ͢Δ CREATE TABLE timeline.user_timeline ( uuid uuid, user_id int, reference
map<text, text>, body text, is_read tinyint, published_at timestamp, PRIMARY KEY (user_id) ) WITH CLUSTERING ORDER BY (published_at DESC);
MATERIALIZED VIEW CREATE MATERIALIZED VIEW timeline.desc_user_timeline AS SELECT uuid, user_id,
published_at, reference, body FROM timeline.user_timeline WHERE user_id IS NOT NULL AND published_at IS NOT NULL AND uuid IS NOT NULL PRIMARY KEY (user_id, published_at, uuid) WITH CLUSTERING ORDER BY (published_at DESC);
From PHP $cluster = Cassandra::cluster() ->withContactPoints('10.0.1.24', ‘localhost') ->withPort(9042) ->build(); $statement
= $session->prepare( "UPDATE users SET age = ? WHERE user_name = ?” ); $futures = array(); // execute all statements in background foreach ($data as $arguments) { $futures[] = $session->executeAsync( $statement, [ ‘arguments' => $arguments ]; }
PHP extension • Batchʹ࠷దԽ͞ΕͨI/F Batch Statement • ฒྻར༻Մೳ • Pagination͕༻ҙ͞Ε͍ͯΔ(Generatorར༻)
• Cassandraͷ΄ͱΜͲͷػೳ͕ར༻Ͱ͖ΔͷͰɺ Java͔Βར༻ͤͣͱे׆༻Ͱ͖Δ
Apache Kafka
Apache Kafka • Streamαϙʔτ(ϥϜμΞʔΩςΫνϟͰඞཁෆՄܽ) • ΫϥελϦϯά͕ࣗ༝ࣗࡏ • Zookeeperͱ࿈ܞͨ͠ࢄγεςϜ • োʹڧ͘ɺϝοηʔδͷ࠶औಘ͕Մೳ
• SparkͱStormͱ༰қʹ࿈ܞͰ͖Δ͜ͱ͔Β࠾༻ • ϝοηʔδૹ৴ޙͰࢦఆظؒอ࣋͠ɺ ଞͷΫΤϦΤϯδϯ͔Βϝοηʔδ༰औಘՄೳ • PHP͔Βར༻Մೳ(rdkafka)
Message QueueͰൃੜ͢Δ • Producer͔ΒBrokerૹ৴࣌ʹܽଛ͢Δ͜ͱ͕͋Δ • Broker͕ड৴Λࣦഊ͢Δέʔε • Brokerͷૹ৴͕ࣦഊ͢Δέʔε • ॏෳͯ͠ड৴ͯ͠͠·͏έʔε
0.11 • Exactly-once delivery and transactional messaging • ਖ਼֬ʹҰ͚ͩɺ࣮֬ʹಧ͚Δ •
ϝοηʔδૹ৴ͱड৴ʹτϥϯβΫγϣϯʂ • ΑΓڧݻʹ
Partition • ฒྻࢄॲཧ͕ઃܭ • topicΛPartitionͰׂ͠ɺProducer, Consumer͕ҙ ͷPartitionʹΞΫηε • ࡉԽͱޮԽ͕ࣗ༝ʹ
Partition
BigDataͷ࢝·ΓPHP͔Β
Presto
େنʹΑΔσʔλநग़ • hdfsʹ֨ೲ͞ΕͨσʔλΛΈ͍ͨ • RDBMSʹ֨ೲ͞Εͨσʔλͱ݁߹ͯ͠΄͍͠ • Ϗδωε؍Ͱͷσʔλूܭநग़Λͯ͠΄͍͠ • σʔλϕʔεࢄͰқ͕ߴ͍ •
खܰʹੳʹར༻͍ͨ͠ -> ແཧ
σʔλϚʔτ • ඞཁͳσʔλΛूΊͯσʔλϕʔεʹू͢Δ • όονॲཧͰ࣮ߦ͢ΔͨΊɺଈ࣌ʹσʔλΛऔಘ͢Δ ͜ͱ͍͠ • σʔλϚʔτࣗମͷอकӡ༻͕ඞཁͱͳΔ (ͦΕ͕ۀͰ͋ΕՄೳ) •
σʔλϚʔτΛઃܭ͢Δͷ͘͠ɺϋʔυϧ͕ߴ͍
Prestoͱ • facebookͰେنͳσʔλʹରͯ͠ɺ ΠϯλϥΫςΟϒʹσʔλऔಘͰ͖ΔΫΤϦΤϯδϯ • ϑϩϯτΞϓϦέʔγϣϯ͔Βhdfsʹଓ͠ɺ σʔλΛଈ࠲ʹՄࢹԽͤ͞Δͷ͍͠ • Hiveόονॲཧ༻్ͷͨΊɺඵͰฦ٫ෆՄೳ (MapReduce)
• RDBMSʹଓ͍ͨ͠ʂͳͲΛղܾ
Prestoͱ • SQLΠϯλʔϑΣʔεΛఏڙ • Cassandra, Hive, Kafka • MongoDB, MySQL,
PostgreSQL, SQLServer • Redis, Thrift • ରԠ͍ͯ͠ͳ͍ͷͰjavaͰυϥΠόΛ࣮͢Δ͚ͩ Ͱ͋Δఔ֦ுՄೳ • SELECTҎ֎ʹINSERTͳͲʹରԠ͓ͯ͠Γɺ σʔλϕʔεҠߦɺࢄΞʔΩςΫνϟͷΧόʔͳͲ ʹ
Prestoͱ
Prestoͱ • jdbcରԠ • PHP͔Βɺ xtendsys-labs/php-presto-client ytake/php-presto-client
None
·ͱΊ • ෳࡶԽ͢ΔΞϓϦέʔγϣϯɺ ՝ղܾେ͖͘ͳΔޣຯ • PHPͰϏδωεΛαϙʔτ͢Δཧπʔϧ • PHP͔Β࢝·ΔBigData + FastDataΞʔΩςΫνϟ
• PHPͰେ͖͘ߩݙ
webΞϓϦέʔγϣϯ͔Β BigData·Ͱࢧ͑ΔPHP