OSC 大阪 2013.cloud - Speaker Deck

Slide 1

Slide 1 text

෼ࢄKVSͱ Ϋϥ΢υετϨʔδ Riak & Riak CS Bashoδϟύϯגࣜձࣾ Senior Software Engineer ্੢߁ଠ 2013/5/25 ΦʔϓϯιʔεΧϯϑΝϨϯε 2013.Cloud @େࡕ

Slide 2

Slide 2 text

Slide 3

Slide 3 text

Slide 4

Slide 4 text

Slide 5

Slide 5 text

Slide 6

Slide 6 text

Slide 7

Slide 7 text

Slide 8

Slide 8 text

©2013 BASHO TECHNOLOGIES INC. ALL RIGHTS RESERVED. Ωʔίϯηϓτ ϚελʔϨε •  ϚϧνϊʔυͷϚελʔϨεΫϥελ KEY VALUE KEY VALUE KEY VALUE bucket •  Key/valueͷϖΞΛbucketsʹอଘ •  ͲͷΑ͏ͳσʔλλΠϓɺΦϒδΣΫτ ΋όΠφϦͱͯ͠อଘ Object/keyΦϖϨʔγϣϯ node node node node node

Slide 9

Slide 9 text

©2013 BASHO TECHNOLOGIES INC. ALL RIGHTS RESERVED. ίϯγεςϯτϋογϡ •  160-‐bit੔਺ΩʔྖҬ •  RingΛ౳ִؒͰύʔςγϣϯʹ෼ׂ •  ύʔςγϣϯΛΫϥελͷ֤ϊʔυ ʹׂΓ౰ͯΔ •  σʔλͷbucket/Keyͷhash஋ʹΑΓ อଘ͢ΔύʔςγϣϯͷҐஔΛܾఆ ͠ɺ࿈ଓ͢ΔύʔςγϣϯʹϨϓϦ έʔγϣϯσʔλΛอଘ node 0 node 1 node 2 node 3 hash(“bucket/key”)

Slide 10

Slide 10 text

Slide 11

Slide 11 text

©2013 BASHO TECHNOLOGIES INC. ALL RIGHTS RESERVED. Ұ࣌ো֐͔Βͷ෮چ࣌ •  Ұ࣌తͳϊʔυো֐ʢnode 2ʣ͕ൃੜ •  PUT, GET, DELETEϦΫΤετ͸ɺϑΥʔϧ όοΫϊʔυʢnode 0ʣ΁ •  ো֐ϊʔυͷ෮چʢnode 2ʣ •  “Handoﬀ”ʹΑΓɺσʔλΛϑΥʔϧόο Ϋϊʔυʢnode 0ʣ͔Β෮چϊʔυ ʢnode 2ʣ΁Ҡߦ •  ௨ৗӡ༻Λ࠶։ node 0 node 1 node 2 node 3 hash(“bucket/key”)

Slide 12

Slide 12 text

©2013 BASHO TECHNOLOGIES INC. ALL RIGHTS RESERVED. •  ಛఆͷbucket / keyΛࢦఆ͠ɺඥ͍ͮͨΦϒδΣΫτΛऔಘ •  ಛఆͷbucketsʹඥ͍͍ͮͯΔશͯͷkeyΛऔಘ Object/KeyΛࢦఆͯ͠ ɹɹΦϒδΣΫτΛऔಘ •  ؆୯ͳΫΤϦݴޠʢHTTPʣΛαϙʔτͨ͠શจݕࡧ •  JSON, plain text, XML and ErlangͳͲͷmime-‐typeΛαϙʔτ •  ·ͩβ yokozuna •  ΫΤϦʔʹ࢖༻Ͱ͖ΔλάɺΛΦϒδΣΫτอଘ࣌ʹઃఆ •  ΦϒδΣΫτͷλά໊͓Αͼ஋Λࢦఆͯ͠ΦϒδΣΫτΛ औಘ Secondary Index (2i) •  σʔλͷ໰͍߹ΘͤɺϑΟϧλϦϯάͷ෼ࢄɺղੳͱूܭ •  Javascript ͱErlang MapReduce Riak಺ͷσʔλ΁ͷΞΫηε

Slide 13

Slide 13 text

©2013 BASHO TECHNOLOGIES INC. ALL RIGHTS RESERVED. ϚϧνσʔληϯλʔϨϓϦέʔγϣϯ ʢMDCϨϓϦέʔγϣϯʣ •  ෳ਺αΠτؒͷσʔλϨϓϦέʔγϣϯ •  σʔλϩʔΧϦςΟ •  ॏେࡂ֐ԼͰ΋αʔϏεΛܧଓ •  ΞΫςΟϒόοΫΞοϓ •  ยํ޲·ͨ͸྆ํ޲ͷಉظ •  ProducUonΫϥελʔͱStage༻Ϋϥελʔ ʹΑΓݕূ؀ڥΛߏங Primary Cluster (DC#1) Secondary Cluster (DC#2) Secondary Cluster (DC#3) Client Update

Slide 14

Slide 14 text

©2013 BASHO TECHNOLOGIES INC. ALL RIGHTS RESERVED. ϨϓϦέʔγϣϯϞʔυ #3 #4,6 Secondary Cluster #3 Primary Cluster #5 #1 TCP connecUon #1) Client͕persistent TCP connecUonΛ։ઃ #2) Client͕Full-‐syncͷىಈΛཁٻ #3) ServerͱClientͷ྆ํͰࣗ਎ͷ࣋ͭobjectͷHashΛܭࢉ #4) Client͕ܭࢉ݁ՌΛServerʹ౉͢ #5) Server͸Client͔Βͷ݁ՌΛൺֱࠩ͠෼Λૹ৴ #6) Client͸ड৴ͨ͠ObjectΛΞοϓσʔτ #6 #3 #4 #1) Primary Cluster͕UpdateϦΫΤετΛड৴ #2) ϨϓϦέʔγϣϯΛ࡞੒ #3) ACKड৴ʹΑΓॻࠐΈ੒ޭͱ൑அ #4) ॻࠐΈ੒ޭޙɺMDC server΁ϦΫΤετૹ৴ #5) Server͸Client΁ObjectΛૹ৴ #6) Secondary ClusterͰϨϓϦέʔγϣϯΛ։࢝ #1 #2 Primary Cluster Secondary Cluster #5 TCP Connection Real-‐Ume sync Full-‐sync Client update

Slide 15

Slide 15 text

Slide 16

Slide 16 text

Slide 17

Slide 17 text

Slide 18

Slide 18 text

©2013 BASHO TECHNOLOGIES INC. ALL RIGHTS RESERVED. ͦͷଞͷػೳ~1.3 •  ΞΫςΟϒɾΞϯνΤϯτϩϐʔʢAAEʣͷ௥Ճ •  ো֐࣌ɺಉ࣌Ξοϓσʔτͦͯ͠෺ཧతͳσʔλܽམ΍ഁյͳͲʹΑΓൃੜͨ͠σʔλͷඇ Ұ؏ੑͷܧଓతͳम෮Λࣗಈతʹ࣮ߦɻ •  ϚϧνσʔληϯλʔϨϓϦέʔγϣϯ΁ͷAdvanced modeͷ௥Ճ •  ৽͘͠Advanced modeʢύΦʔϚϯε޲্ɺෳ਺TCPίωΫγϣϯͱ؆୯ͳઃఆʣΛ௥Ճɻ •  IPv6αϙʔτͷ֦ு

Slide 19

Slide 19 text

©2013 BASHO TECHNOLOGIES INC. ALL RIGHTS RESERVED. Riak1.4ͷ৽نػೳ(2013.June?) •  ϦϯάαΠζͷมߋ •  ༰ྔޮ཰ͷ޲্ •  2i Ͱͷ໰͍߹Θͤޮ཰Խ •  MDCϨϓϦέʔγϣϯAdvanced mode ʹSSLɼNAT, ϓϩΩγ௥Ճ •  Riak ControlͷGUIվળ What’s new on Riak1.4

Slide 20

Slide 20 text

Slide 21

Slide 21 text

©2013 BASHO TECHNOLOGIES INC. ALL RIGHTS RESERVED. ϊʔυ௥Ճ 1.  ௥Ճ͢ΔϚγϯΛىಈ 2. $ riak-admin cluster \! join riak@10.0.0.1! 1. $ riak-admin cluster plan! 2. $ riak-admin cluster commit! ͢Ͱʹىಈͯ͠ ͍ΔΫϥελͷ ͏ͪ୭͔

Slide 22

Slide 22 text

Slide 23

Slide 23 text

Slide 24

Slide 24 text

Slide 25

Slide 25 text

Slide 26

Slide 26 text

Slide 27

Slide 27 text

Slide 28

Slide 28 text

Slide 29

Slide 29 text

©2013 BASHO TECHNOLOGIES INC. ALL RIGHTS RESERVED. APIͱΠϯλϑΣʔε •  ϢʔβΞΧ΢ϯτ࡞੒ɺΞΧ΢ϯτແ ޮԽͱ؅ཧ༻ͷΠϯλϑΣʔε •  ඪ४ͷS3ೝূεΩʔϚʢϔομͱΫΤ ϦʔύϥϝʔλೝূΛαϙʔτʣ •  ؅ཧऀͷΈ͕ϢʔβΞΧ΢ϯτͷ࡞੒ ΛͰ͖ΔΑ͏ʹઃఆՄೳ AdministraUon, Users and AuthenUcaUon

Slide 30

Slide 30 text

©2013 BASHO TECHNOLOGIES INC. ALL RIGHTS RESERVED. ϚϧνσʔληϯλʔϨϓϦέʔγϣϯ •  ϢʔβؒͰάϩʔόϧͳ৘ใʢόέοτͱ ϚχϑΣετʣ͸ɺϦΞϧλΠϜʹϨϓϦ έʔγϣϯ •  ΦϒδΣΫτ৘ใͷϨϓϦέʔγϣϯ͸ɺ ϑϧಉظ(Full-‐sync)·ͨ͸ϦΞϧλΠϜಉظ (Real Ume)Λબ୒ •  Ϣʔβ͕ΦϒδΣΫτͷGETΛཁٻͨ͠ࡍ ʹɺΞΫηεͨ͠αΠτʹ͸ΦϒδΣΫτ Λߏ੒͢ΔϒϩοΫͷҰ෦͕ϨϓϦέʔγ ϣϯ͞Ε͍ͯͳ͔ͬͨ৔߹ɺ଍Γͳ͍ϒϩ οΫΛ”ΦϦδϯ”ͷΫϥελʔଆ͔Βऔಘɺ ࠶ߏ੒ How It Works

Slide 31

Slide 31 text

Slide 32

Slide 32 text

Slide 33

Slide 33 text

©2013 BASHO TECHNOLOGIES INC. ALL RIGHTS RESERVED. Riak ͷద༻γφϦΦ •  Ad Networks •  Digital Media •  On-‐Line Games •  Social Networks •  Social Analysis •  Cloud Operators •  Messaging Services •  Product Catalogs •  Document Management •  Health Care InformaUon Management

Slide 34

Slide 34 text

Slide 35

Slide 35 text

©2013 BASHO TECHNOLOGIES INC. ALL RIGHTS RESERVED. ੡඼৘ใͷอ؅ •  ֦ுੑͱϢʔβΤΫεϖϦΤϯεͷύϑΥ ʔϚϯεվળ •  bestbuy.comͱখചΓళฯͰ࢖༻͢ΔΦϯ ϥΠϯ੡඼ΧλϩάɺϨʔςΟϯάʹRiak Λબ୒ɻ •  Holiday ShoppingʢΫϦεϚεηʔϧͳ Ͳʣ࣌ʹ߹Θͤͯϊʔυ௥Ճɻ •  Bestbuy.comͷϗʔϜϖʔδϨϯμϦϯά ͷSLA͸̍ඵҎ಺ɻ •  2013೥ͷHoliday seasonʹ͸ɺSKU਺Ͱ26ඦ ສ͔Β500ඦສ΁֦ுΛ૝ఆɻ •  Amazon AWSͷෳ਺Availability Zone্ʹRiak ΫϥελʔΛߏங

Slide 36

Slide 36 text

©2013 BASHO TECHNOLOGIES INC. ALL RIGHTS RESERVED. ϝχϡʔγεςϜ •  ౰ॳAmazon S3Λ࢖༻; On demandϝχϡʔͷϥ΢ϯ υτϦοϓϨΠςϯγʔͷ௿ݮΛୡ੒ •  ߴ଎ͳಡΈग़͠/ॻࠐΈͷͨΊRiakΛબ୒ •  Video On demandϝχϡʔͷΞάϦήʔτʹඞཁͳ σʔλΛอଘ •  ϚʔέςΟϯάΩϟϯϖʔϯʹকདྷ࢖༻͢ΔͰ͋Ζ ͏Ϣʔβؔ࿈ͷ৘ใΛอଘ •  ຖ೔ɺϦϞʔτίϯτϩʔϥ͔Βͷ25MΫϦοΫ਺ Λॲཧ •  On demandϝχϡʔͷͨΊʹ̏ΧॴͷDCʹRiakΫϥ ελʔΛߏஙɻ •  ϚʔέςΟϯά༻ʹ̐ͭͷΫϥελʔΛߏங

Slide 37

Slide 37 text

©2013 BASHO TECHNOLOGIES INC. ALL RIGHTS RESERVED. ϔϧεέΞ৘ใ؅ཧ •  HealthσʔλΛγΣΞɺΞΫηε͢ΔͨΊͷHIPAA४ ڌͷWebͱܞଳి࿩ͷϓϥοτϑΥʔϜΛߏங •  ίετ௿ݮͱߴՄ༻ੑɾ֦ுੑͷཱ྆ •  ߴ଎ͳಡΈग़͠/ॻࠐΈͷͨΊRiakΛબ୒ •  mHealth͸ɺܞଳి࿩ɺαʔϏε͓Αͼҩྍػث͔ ΒͷσʔλΛऩूʢױऀ͕औಘσʔλΛঝೝʣ •  2015೥ʹ500M appsΛ૝ఆɻ •  ̏ΧॴͷDCʹRiakΫϥελʔΛߏஙɻ

Slide 38

Slide 38 text

©2013 BASHO TECHNOLOGIES INC. ALL RIGHTS RESERVED. Πϕϯτ؅ཧ •  ϥΠηϯείετͷ௿ݮ͓Αͼӡ༻ͷ༰қ͞ •  Ϣʔβొ࿥৘ใ͓ΑͼϢʔβΠϕϯτϩάͷอଘɻ ʢϢʔβ͕Webར༻ঢ়گϞχλʔΛޙ͔ΒνΣο Ϋʣ •  Πϕϯτ௨஌αʔϏεʹͯ৽ن৘ใΞοϓσʔτ௨ ஌Λ഑৴͢ΔͨΊͷΫΤϦ༻DBʹ࢖༻ •  14,000 read request/sec •  2013೥தʹ3ՕॴͷDCϔ֦ு༧ఆ •  Πϕϯτ௨஌αʔϏε͸ɺ50ඦສίωΫγϣϯ/೔ Λ૝ఆɻ

Slide 39

Slide 39 text

©2013 BASHO TECHNOLOGIES INC. ALL RIGHTS RESERVED. ΦϯϥΠϯ޿ࠂ •  DCϨϓϦέʔγϣϯͱ֦ுੑͷඞཁ͔Β MySQL → Cassandra → Riak ͱϦϓϨʔε •  ϢʔβΞΫςΟϏςΟσʔλ͓ΑͼτϥϑΟο ΫσʔλΛอଘ •  τϥϑΟοΫσʔλ͸MySQL͔ΒҠߦʢෳ਺DC ؒͷσʔλϨϓϦέʔγϣϯ͕ඞཁʣ •  Ϣʔβσʔλ͸Cassandra͔ΒҠߦʢbackward compaUbility͕֬อͰ͖͍ͯͳ͍͜ͱ͕աڈͷ ϦϦʔεʹ͋Γɺ৴༻Ͱ͖ͳ͍ͨΊܾஅʣ •  5ΧॴͷDCؒͰσʔλϨϓϦέʔγϣϯ •  2011೥ʹ4 trillionͷadσʔλΛѻ͏ɻ advertising conomic potential of uding OpenX mpTime) provide a bining ad serving, an ad exchange, a Supply Side Platform, and ach year. OpenX uses Riak for user and trafficking data behind its data services API. They selected Riak due to its highly available, low-latency, redundant architecture. OpenX also uses Riak’s multi- datacenter replication across several data centers, providing up-to-date data throughout its global infrastructure. For more details about how OpenX uses Riak, check out the video of Anthony Molinaro, OpenX engineer, speaking at RICON2012, Basho’s 2012 developer conference. ng technology provider. gencies, mobile operators, active and measurable obile devices. In 2009, rly all of the broadcasters le operators. With the dly, they needed to move to an architecture that could gracefully new platform because it is distributed, scalable, and highly mes of traffic. they opted to build two geographically separated, mirrored sites ation feature. As Marcus Kern, VP of Technology at Velti, ver 140 customers. We cannot afford a single minute of d exceed our requirements for scale, data durability, and 2009೥࣌఺

Slide 40

Slide 40 text

©2013 BASHO TECHNOLOGIES INC. ALL RIGHTS RESERVED. υΩϡϝϯτετϨʔδ •  ΦϒδΣΫτετΞΛMySQL͔ΒRiak΁Ϧϓ Ϩʔε •  ΞϓϦέʔγϣϯ͕ѻ͑ΔΦϒδΣΫτϑΝ ΠϧͷαΠζ֦େԽΛ࣮ࢪ •  HRΞϓϦέʔγϣϯ͕ऩू͢ΔΦϒδΣΫτ ϑΝΠϧΛอଘ •  ΦϒδΣΫτϑΝΠϧΛෳ਺ͷDCͷΫϥελ ʹϨϓϦέʔγϣϯɻ •  ·ͨQA΍࣍ظϦϦʔεͷͨΊͷStage༻Ϋϥε λ΁΋ϨϓϦέʔγϣϯɻ •  ̕ΧॴͷΫϥελʢStage༻ΫϥελΛؚΉʣ Workday DRIVERS 1.  Replace MySQL object store. 2.  Implement ability to dramatically increase the size of attachments accepted by the Workday application. STATS 1.  Nine 5 node clusters spread across 3 data centers and 3 application zones. FACTOIDS 1.  At Workday IPO in October 2012, Workday shares opened at $48.05 USE CASE 1.  Host and store attachments collected by the Workday HR application. 2.  Replicate attachments across several clusters including initial customer setup, ongoing production, and QA for upcoming releases. Workday Application! Riak! Workday Blobitory! Real Ume replicaUon Prod#1 (Dublin) Prod#2 (Ams) Stage (Dublin) BidirecUonal replicaUon Real Ume DirecUonal ReplicaUon

Slide 41

Slide 41 text

©2013 BASHO TECHNOLOGIES INC. ALL RIGHTS RESERVED. :BIPP+"1"/ •  ΠϯλʔωοτγϣοϐϯάαΠτͷϓ ϥοτϑΥʔϜΛYahoo!δϟύϯ͕ఏڙ •  γϣοϐϯάαΠτͷը૾σʔλΛRiak CSʹετΞ •  ొ࿥ΦϒδΣΫτ਺ɿ20ສ݅ʢ2012೥຤ ࣌఺ʣ •  ϦΫΤετ਺ɿ450 req/sec •  Ϩεϙϯεɿ10ms – 80ms •  ߏஙɿ1೔ɻ •  S3ޓ׵Ϋϥ΢υετϨʔδαʔϏεΛఏڙ •  2ΧॴͷDCؒͰσʔλϨϓϦέʔγϣϯ

Slide 42

Slide 42 text