セプテーニで分析基盤(Treasure Data)を導入した話

新宿Geek Lounge#3 分析基盤Meetup　登壇資料

Transcript

1. @৽॓Geek Lounge#3 ෼ੳج൫Meetup
   @kimutyam
   2017/09/19
   ηϓςʔχͰ෼ੳج൫
   (Treasure Data)Λಋೖͨ͠࿩
   

   View Slide

2. ໊લ໦ଜজ޺
   5XJUUFS!LJNVUZBN
   ձࣾηϓςʔχɾΦϦδφϧ
   %%%ͱ*$0/*9Λ࢖ͬͨ෼ੳઃܭΛ޷Έͱ͍ͯ͠Δ4DBMBΤϯδχΞ
   ۀ຿ࢧԉͷΞʔΩςΫνϟΛߟҊ͢Δ͏ͪʹ෼ੳج൫ܥͷۀ຿ʹ୧Γண͖·ͨ͠
   ࣗݾ঺հ
   

   View Slide

3. PYXISͷྺ࢙ͱ෼ੳج൫ಋೖ·Ͱͷഎܠ
   

   View Slide

4. PYXISͱ͸
   ޿ࠂ഑৴
   ࠷దԽ
   ୅ཧళۀ
   ຿վળ
   σʔλ෼
   ੳ
   

   View Slide

5. σʔλ෼ੳॳظ
   PYXISͷσʔλ ηϓςʔχͷ֤෦໳DB
   ഔମࣾͷσʔλ
   ސ٬ͷσʔλ
   ࣾһͷσʔλ
   ֎෦ϕϯμʔͷσʔλ
   ಛఆͷ໨తʹԊͬͯߏ଄Խ͞Εͨσʔλ͕෼ࢄ؅ཧ͞Εɺ࣍ར༻͞Ε͍ͯͨ
   

   View Slide

6. σʔλͷ࣍ར༻͕ͮ͠Β͍
   σʔλ෼ੳॳظͰͷ໰୊఺
   ύϑΥʔϚϯε͕ѱ͍
   σʔλ͕෼ࢄ؅ཧ͞Ε͍ͯΔ
   &5-ॲཧ"11ͷཚཱ
   

   View Slide

7. σʔλͷ࣍ར༻͕ͮ͠Β͍
   վળ΁
   ύϑΥʔϚϯε͕ѱ͍ &5-ॲཧ"11ͷཚཱ
   ඇߏ଄Խσʔλͷ஝ੵ
   ෼ࢄॲཧ "11ͷ؆ུԽπʔϧར༻
   σʔλͷू໿Խ
   σʔλ͕෼ࢄ؅ཧ͞Ε͍ͯΔ
   

   View Slide

8. σʔλͷ࣍ར༻͕ͮ͠Β͍
   ιϦϡʔγϣϯ: σʔλϨΠΫ
   ύϑΥʔϚϯε͕ѱ͍ &5-ॲཧ"11ͷཚཱ
   ඇߏ଄Խσʔλͷ஝ੵ
   ෼ࢄॲཧ "11ͷ؆ུԽπʔϧར༻
   σʔλͷू໿Խ
   σʔλ͕෼ࢄ؅ཧ͞Ε͍ͯΔ
   

   View Slide

9. σʔλͷ࣍ར༻͕ͮ͠Β͍
   ιϦϡʔγϣϯ: σʔλ΢ΣΞϋ΢ε
   ύϑΥʔϚϯε͕ѱ͍
   σʔλ͕෼ࢄ؅ཧ͞Ε͍ͯΔ
   &5-ॲཧ"11ͷཚཱ
   ඇߏ଄Խσʔλͷ஝ੵ
   ෼ࢄॲཧ "11ͷ؆ུԽπʔϧར༻
   σʔλͷू໿Խ
   

   View Slide

10. σʔλͷ࣍ར༻͕ͮ͠Β͍
    ιϦϡʔγϣϯ: Treasure Data
    ύϑΥʔϚϯε͕ѱ͍
    σʔλ͕෼ࢄ؅ཧ͞Ε͍ͯΔ
    &5-ॲཧ"11ͷཚཱ
    ඇߏ଄Խσʔλͷ஝ੵ
    ෼ࢄॲཧ "11ͷ؆ུԽπʔϧར༻
    σʔλͷू໿Խ
    

    View Slide

11. Treasure DataΛར༻ͨ͠ΞʔΩςΫνϟ
    

    View Slide

12. ΞʔΩςΫνϟߏ੒
    1MB[NB%# %BUB5BOL
    

    View Slide

13. S3Λ1࣍੥͚ʹ͢Δ
    ετϨʔδ͸ͳ͍ͷͰ4Λར༻
    ੜσʔλͷσʔλܗࣜ +40/$479.-ͳͲ
    ͷ··อଘ
    4ϑΝʔετʹ͢Δ͜ͱͰ"UIFOB3FETIJGU&.3ͳͲͷ
    αʔϏεͱͷ࿈ܞ͕ඞཁʹԠͯ͡Մೳʹ
    ੜσʔλอଘͷࡍʹσʔλεΩʔϚΛߟྀ͠ͳ͍
    

    View Slide

14. ߏ଄Խσʔλ͸ڥք͚ͮΒΕͨίϯςΩετʹด͡Δ
    ޿ࠂ഑৴
    ࠷దԽ
    ୅ཧళۀ
    ຿վળ
    σʔλ෼
    ੳ
    ʮ୅ཧళۀ຿վળʯ΍ʮ޿ࠂ഑৴࠷దԽʯͷίϯς
    Ωετ಺ʹσʔλ͸1:9*4಺ʹݶఆ
    ʮσʔλ෼ੳʯͷίϯςΩετͰ͸࣍ར༻͢Δͨ
    ΊʹੜσʔλΛ4ʹอଘ
    

    View Slide

15. S3ͷϨΪϡϨʔγϣϯ
    #VDLFU1:9*4ͷΞϓϦέʔγϣϯڥքͷ໋໊ʹ͢Δ
    1BUI\5%ͷσʔλϕʔε໊^\5%ͷςʔϒϧ໊^ZZZZ..ZZZZ..EE@
    ΞϓϦέʔγϣϯؒͷڝ߹Λճආ͢ΔͨΊ
    &NCVMLଆͰQBUI@QSFpYͰ೔࣌ΛಈతʹࢦఆͰ͖ΔΑ͏ʹ
    "UIFOBͷύʔςΟγϣϯͰ೔෇ࢦఆ͠΍͘͢ͳΔΑ͏ʹ
    
    

    View Slide

16. S3 -> TD
    

    View Slide

17. Treasure DataͷྖҬ
    1MB[NB%# %BUB5BOL
    5SFBTVSF%BUBͷϑϧϚωʔδͷαʔϏεΛར༻
    5SFBTVSF8PSLqPX %JHEBH
    
    %BUB$POOFDUPS &NCVML
    
    5SFBTVSF8PSLqPXΛར༻͢Δ͜ͱͰɺಠࣗʹ%JHEBH
    4FSWFSΛཱͯΔߏஙɾอकίετΛ࡟ݮ
    5SFBTVSF8PSLqPX͔Βར༻͠ɺඞཁʹԠͯ͡%JHEBH
    4FSWFSΛཱͯΔΑ͏ʹϓϩδΣΫτελʔτ
    

    View Slide

18. Treasure DataͷྖҬ
    1MB[NB%# %BUB5BOL
    4͔ΒUE@MPBE
    1MB[NB%#͔Β1SFTUPͰूܭͯ͠൚ԽσʔλΛ1MB[NB%#
    ʹอଘ
    %BUB5BOL σʔλϚʔτ
    ʹར༻༻్ʹԊͬͨಛԽ͞Εͨ
    σʔλΛ֨ೲ
    ˞%BUB5BOL͸5SFBTVSF%BUB͕ఏڙ͍ͯ͠Δ1PTUHSFTRM
    

    View Slide

19. ࣗલͰDigdag ServerΛݕ౼ͨ͠ྫ
    1MB[NB%# %BUB5BOL
    42-͚ͩͰ͸ूܭग़དྷͳ͍৔߹
    %JHEBHͷQZUIPOSVCZTIFMMΦϖϨʔλʔΛར༻
    ͯ͠ɺಠࣗʹ࣮૷ 5SFBTVSF8PSLqPX͸ະରԠ
    
    %BUB$POOFDUPSͰ͸ཉ͍͠ूܭཻ౓͕ಘΒΕͳ͍৔߹
    &NCVMLϓϥάΠϯΛ૊ΈࠐΈPS։ൃ
    

    View Slide

20. Treasure DataΛར༻ͨ͠ۀ຿ϑϩʔ
    σʔλ஝ੵ
    σʔλ෼ੳɾՁ஋ݕূ
    %JHEBHʹΑΔΦʔτϝʔγϣϯԽ
    ։ൃऀ͕5SFBTVSF%BUBʹσʔλ֨ೲ
    σʔλΞφϦετ ฐࣾͰ͸ඇΤϯδχΞ͕ଟ͍
    ͕5SFBTVSF%BUBͷ(6*Λར༻
    ։ൃऀ͕؅ཧ͢Δ
    5SFBTVSF%BUBΛհͯ͠։ൃऀͱσʔλΞφϦετͷۀ຿࿈ܞΛ͢Δ
    

    View Slide

21. Treasure DataΛબ୒ͨ͠Α͔ͬͨ఺
    

    View Slide

22. ϑϧϚωʔδυͳ෼ੳج൫ΛεϐʔσΟʔʹखʹೖΓɺ෼ੳۀ຿ʹूத͕Ͱ͖Δ
    "84΍($1Ͱ෼ੳج൫Λࣗલߏங͢Δ৔߹͸ɺ
    αʔϏεͷ஌ݟΛ্࣋ͬͨͰอकɾӡ༻Λߦ͏ඞཁ͕͋Δ
    ˞ಋೖޙʹඇػೳཁٻ͕໌֬Խͨ͠Βࣗલ΋ݕ౼΋
    ෼ੳۀ຿ʹूதͰ͖Δ
    

    View Slide

23. σʔλͷΠϯϙʔτΤΫεϙʔτͷΤίγεςϜͷॆ࣮͍ͯ͠·͢
    ֤छσʔλϕʔε΍ϑΝΠϧαʔόʔͱͷૄ௨ͷόϦΤʔγϣϯ͕ଟ͘ɺ
    ਺ଟ͘ͷ֎෦γεςϜͱͷ࿈ܞΛ༰қͰ͢
    ֎෦γεςϜͱͷ࿈ܞ͕༰қ
    

    View Slide

24. 5SFBTVSF%BUBʹσʔλ஝ੵΛ͢Ε͹ɺσʔλΞφϦετ͕෼ੳͰ͖Δঢ়ଶʹ ۃ࿦
    
    ΫΤϦ՝ۚ΍సૹྔ՝ۚͰ͸ͳ͍ͷͰ42-ʹਫ਼௨͍ͯ͠ͳ͍͕ΫΤϦൃߦͯ͠΋Ϋϥ΢υഁ࢈͠ͳ
    ͍
    σʔλ෼ੳɾՁ஋ݕূ·Ͱͷεϐʔυײͱ҆৺ײ
    

    View Slide

25. ͝੩ௌ͋Γ͕ͱ͏͍͟͝·ͨ͠
    

    View Slide