Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Speaker Deck
PRO
Sign in
Sign up
for free
大規模Webサービス入門 5回目 / Introduction to large scale web service 5
muttan
August 11, 2017
Technology
0
43
大規模Webサービス入門 5回目 / Introduction to large scale web service 5
muttan
August 11, 2017
Tweet
Share
More Decks by muttan
See All by muttan
bath_poo_
0
22
bath_poo_
0
71
bath_poo_
0
220
bath_poo_
0
80
bath_poo_
0
80
bath_poo_
1
170
bath_poo_
0
80
bath_poo_
0
110
bath_poo_
0
160
Other Decks in Technology
See All in Technology
meteatamel
0
440
oyakata2438
0
410
hololab
0
360
becolomochi
0
170
kanaugust
PRO
0
120
qryuu
7
5k
mitsu9
0
130
muras
0
120
yamasaki696
0
400
takaking22
2
530
kraj
0
4.1k
clustervr
0
170
Featured
See All Featured
morganepeng
92
14k
erikaheidi
13
4.2k
kneath
294
39k
edds
56
9.3k
nonsquared
81
3.3k
shpigford
165
19k
sachag
267
17k
jonyablonski
14
1.1k
kneath
219
15k
searls
204
35k
eitanlees
111
9.9k
brad_frost
157
6.4k
Transcript
େنαʔϏεٕज़ೖ ୈ5ճ ISUCONରࡦษڧձ 2017/8/11
ୈ5ճ େنσʔλॲཧ[࣮ફ]ೖ - ΞϓϦέʔγϣϯ։ൃͷצॴ -
େنσʔλॲཧΞϓϦέʔγϣϯͷ ߟ͑ํͱରࡦ • ࠓ·Ͱ, େྔͷσʔλ͕͋ͬͯϋʔυΣ ΞͷߏΛݟ͢ʢہॴੑΛ׆͔͢ʣ͜ͱͰ ͳΜͱ͔ΓΖ͏ͱ͍͏ߟ͑. • Ͳ͏ͯ͠େྔͷσʔλʹΞΫηε͠ͳ͍ͱ ͍͚ͳ͍ͱ͖ͷରࡦΛߟ͑Δ.
େنσʔλॲཧΞϓϦέʔγϣϯͷ ߟ͑ํͱରࡦ • ຊͷ༰ • Lesson14 ༻్ಛԽܕΠϯσΫγϯά • Lesson15 ཧͱ࣮ફͷ྆ଆ͔Β߈ΊΔ
Lesson14 ༻్ಛԽܕΠϯσΫγϯά
ΠϯσοΫεͱγεςϜߏ • େنσʔλΛѻ͏ྫ • શจݕࡧ • ྨࣅจॻܥ୳ࡧ • σʔλϚΠχϯά
ΠϯσοΫεͱγεςϜߏ • ઌఔڍ͛ͨγεςϜͰRDBMSͰ͠ΜͲ͍ • ͳΒRDBMSΛΘͳ͚Ε͍͍͡Όͳ͍ • શ͘RDBMSΛΘͳ͍ͱ͍͏Ͱͳ͍
ΠϯσοΫεͱγεςϜߏ • σʔλRDBMSʹอଘ͓ͯ͘͠ • ͦͷσʔλΛఆظతʹநग़͠, ผ్ΠϯσοΫ εαʔόͷΑ͏ͳͷΛ࡞Δ • ͦ͜ʹWebΞϓϦ͔ΒRPCͳͲͰΞΫηε͢ Δํ๏Λ͏
ʲ෮शʳRDBMS • Relational DataBase Management System • ؔσʔλϕʔεΛӡ༻͢ΔͨΊͷιϑτΣ Ξͷू߹ମ •
MySQL, PostgreSQL, SQLite, etc…
ʲ෮शʳRPC • Remote Procedure Call • ωοτϫʔΫʹଓ͞Εͨଞͷίϯϐϡʔλ ্ͷϓϩάϥϜΛ࣮ߦ͢Δ • ͦͷ݁ՌΛडऔΔ
ΠϯσοΫεͱγεςϜߏ ΠϯσοΫε αʔό ᶃఆظతʹ σʔλΛநग़ cron job NPE@QFSM "1TFSWFS ᶄΠϯσοΫεΛ࡞Δ
ᶅRPCͰΞΫηε ᶆσʔλऔಘ
ΠϯσοΫεͱγεςϜߏ2 ΠϯσοΫε αʔό ᶃఆظతʹ σʔλΛநग़ cron job NPE@QFSM "1TFSWFS ᶄΠϯσοΫεΛ࡞Δ
ᶅRPCͰΞΫηε ᶇσʔλऔಘ httpd ᶆݕࡧͯ͠JSON Λฦ͢ΞϓϦ
ͳͥAPαʔόʹΠϯσοΫεΛ ࣋ͨͤͳ͍ͷ͔ • ઌ΄Ͳͷਤͷmod_perlͷ෦ • ेͳϝϞϦ͕ͳ͍ • ΞʔΩςΫνϟతʹେྔͷσʔλΛ୳͢Α͏ ͳͷʹ͍͍ͯͳ͍
ͳͥAPαʔόʹΠϯσοΫεΛ ࣋ͨͤͳ͍ͷ͔ • APαʔόશͯʹΠϯσοΫεΛͨͤΔͷ େมʢࠓޙ૿͑ΔՄೳੑ͕͋Δʣ • ΠϯσοΫεαʔόʹूͯ͠ཧ͢Δ
RPCͬͯͬͯΔʁ • ͔ͭͯRPCΛͬͯΠϯσοΫεαʔό͔Β ݁ՌΛऔಘ͍ͯͨ͠.ʢࠓ͋Δʣ • ࠓͰJSON+HTTP͕ओྲྀ • զʑೃછΈ͕ਂ͍
༻్ಛԽܕͷΠϯσΫγϯά • ࠓ·Ͱհͨ͠ߏʮ༻్ಛԽܕΠϯσΫ γϯάʯͱݺΕ͍ͯΔ • ͜ΕʹΑͬͯ, RDBMSͰ͔ͬͨ͜͠ͱ͕ ࣮ݱՄೳʹʂ
༻్ಛԽܕͷΠϯσΫγϯά • RDBMS൚༻తʹ͑ΔΑ͏༷ʑͳػೳ͕උ Θ͍ͬͯΔ • ౷ܭॲཧ, ݁߹, ιʔτ • ཉ͍͠ػೳ͚ͩʹಛԽʢνϡʔχϯάʣ͢Δ
ͨΊૣ͘ͳΔ
༻్ಛԽܕͷΠϯσΫγϯά • σʔλΛఆظతʹॻ͖ग़ͯ͠ΠϯσοΫεʢσʔλ ߏʣΛߏங͢Δ • ߏԽͨ͠σʔλΛอ࣋ͨ͠αʔόΛC++Ͱ࡞Γ, RPCͰΞΫηε͢ΔͳͲ • ThriftͬͯͷͰଟݴޠRPC͕Մೳʹ •
ௐͯΈΔͱݕࡧΤϯδϯܥͰΑ͋͘ΔߏͬΆ͍
ʲྫʳͯͳΩʔϫʔυʹΑΔϦϯΫ
ʲྫʳͯͳΩʔϫʔυʹΑΔϦϯΫ • ΩʔϫʔυϦϯΫΛੜ͢ΔॲཧΛߟ͑Δ Լઢ෦͕ϦϯΫ
ʲྫʳͯͳΩʔϫʔυʹΑΔϦϯΫ • Ωʔϫʔυͷ͕20ສϫʔυڧ͋Δ • ͜ΕΛ͍͍ͪͪൺֱ͍ͯ͘͠ͱաෛՙʹͳͬ ͯDBαʔό͕མͪͯ͠·͏ • Ͳ͏ͨ͠Β͍͍ͩΖ͏͔ʁ
ʲྫʳͯͳΩʔϫʔυʹΑΔϦϯΫ • લʹͬͨΑ͏ʹ, όονॲཧͰΩʔϫʔυΛ ͯ͢औΓग़͓ͯ͘͠ • ͔ͭͯڊେͳਖ਼نදݱΛ༻͍ͯνΣοΫ͠ ͍ͯͨ • OR݅ʹΑΓόοΫτϥοΫ͕ͨ͘͞Μൃ
ੜͯ͘͠ͳͬͨ
ʲྫʳͯͳΩʔϫʔυʹΑΔϦϯΫ • ݱࡏCommon Prefix Searchʢڞ௨಄ࣙݕࡧʣͱ TrieΛͬͯϚονϯά͍ͯ͠Δ • Common Prefix Searchʹ,
Aho-Corasick๏ʢΤΠ ϗʔίϥγοΫʣDouble Array TrieͳͲ • ࣗવݴޠॲཧാͩͱԦಓͳํ๏Β͍͠ • Aho-Corasickؤுͬͯௐ͍ͯͩ͘͞
Trieͱ • Ωʔू߹Λѻ͏ͨΊͷσʔλߏͷҰछ • ࠓճͷΑ͏ͳ୯ޠͷू߹ͱ͔ • ݕࡧαΠζ͕ͷେ͖͞Ͱͳ͘୯ޠͷ͞ ʹґଘ͢Δ • ऩ݅ʹґଘ͠ͳ͍
Trieͱ t e a n o i n n w
e keys: tea, ten, to, i, in, inn, we
ʲྫʳͯͳΩʔϫʔυʹΑΔϦϯΫ • հͨ͠Α͏ͳσʔλߏΛ༧Ίߏங͓ͯ͠ ͘͜ͱͰߴԽग़དྷΔ • 7ճͰৄ͘͠આ໌
ʲྫʳͯͳϒοΫϚʔΫͷ ςΩετྨث
ʲྫʳͯͳϒοΫϚʔΫͷ ςΩετྨث • ͯͳϒοΫϚʔΫͷΧςΰϦࣗಈྨ, Complement Naive Bayesͱ͍͏ΞϧΰϦζϜ ΛͬͯػցֶशΛߦ͍ྨ͍ͯ͠Δ. • ৄׂ͘͠Ѫ͢Δ͕,
ग़ݱසΛසൟʹٻΊΔ ͜ͱʹͳΔͷͰ, ͦΕ͚ͩΛฦ͢αʔό͕ଘࡏ ͍ͯ͠Δ
Lesson15 ཧͱ࣮ફͷ྆ํ͔ΒऔΓΉ
ٻΊΒΕΔٕज़తͳཁ݅ΛݟۃΊΔ • ཧ • ίϨΛ͜ͷ༷ʹ͢Δͱಈ͘Αͱ͍ͬͨΑ͏ͳ ࣝ • ࣮ફ • ࣮ࡍʹखΛಈ͔ͨ͠Γӡ༻্͍ͯ͘͠Ͱඞཁ
ʹͳͬͯ͘Δϊϋ
େنΞϓϦʹ͓͚Δཧͱ࣮ફ • ཧͱ࣮ફΛόϥϯεྑ͘Βͳ͍ͱμϝ • ͲͪΒ͔Ұํ͚ͩʹภ͍͚ͬͯͳ͍ • ཧ͚͍͍࣮ͩͬͯͯ͟ग़དྷΔ͔ͱ͍͏ͱ, ࣮ ͷͨΊͷόουϊϋ͕… •
࣮ફΛ͜ͳ͍ͯͯ͠, Θ͔Βͳ͍͜ͱʢࣝෆʣ Ͱͯ͘Δ
ܭࢉػͷͱͯ͠ಓےΛݟ͚ͭΔ • ઌͷΩʔϫʔυΛݟ͚ͭΔॲཧͰ, Double array Trieͱ͍͏͕ग़͖ͯͨ • ͜ͷΑ͏ͳΞϧΰϦζϜΛ͍ͬͯΔ͔Β ͳ͍͔Ͱେ͖͘มΘͬͯ͘Δ •
ʲ࠶ܝʳΞϧΰϦζϜେࣄ
2ճ͔Β5ճ·Ͱͷখ·ͱΊ
ୈ2ճʙୈ5ճͷখ·ͱΊ 1. ΪΨόΠτ୯Ґͷσʔλॲཧ ςϥ, ϖλόΠτͷσʔλΛѻ͏ʹͲ͏͢Δ͔. 2. ϝϞϦॏཁ ϝϞϦʹࡌΔͳΒϝϞϦʹ. Ωϟογϡ͕ฉ͖͍͢ߏʹ͢Δ. 3.
ࢄΛҙࣝͨ͠ӡ༻ దͳεΩʔϚͷઃఆ, ύʔςΟγϣχϯά, JOINΛආ͚Δ. 4. దͳΞϧΰϦζϜͱσʔλߏ Trie, Double Array Trie, Common Prefix Search