Tweet
Tweet

Slide 1

Slide 1 text

೔ܦͷσʔλϕʔεࣄۀͱ Elasticsearch ೔ᙛ ହี ݕࡧΤϯδχΞ @ Nikkei Inc. 2025/02/26 Elasticsearch Community in Tokyo 2025

Slide 2

Slide 2 text

೔ܦͷσʔλϕʔεࣄۀ

Slide 3

Slide 3 text

σʔλͰ ݟ௨͢ɺܾΊΔɺಈ͘ɻ త֬ͳҙࢥܾఆΛɻ https://nkbb.nikkei.co.jp/about/

Slide 4

Slide 4 text

೔ܦͷσʔλϕʔεࣄۀ શίϯςϯπ ݕࡧج൫ (CPF) 200 ࣾҎ্͔Βίϯςϯπఏܞ ɾ ɾ ɾ ɾ ɾ ɾ 12 ͷ৘ใαʔϏε +

Slide 5

Slide 5 text

● ࠃ಺࠷େڃͷϏδωεσʔλϕʔεαʔϏε ○ 750Ҏ্ͷ৘ใഔମ ○ 8,600ສࣾͷاۀ৘ใ ○ 30ສͷਓࣄ৘ใ ● ౦ূ1෦্৔اۀͷ7ׂҎ্͕ಋೖ ● શจݕࡧͰҙࢥܾఆΛࢧԉ

Slide 6

Slide 6 text

ݕࡧٕज़͕ࢧ͑Δσʔλϕʔεࣄۀ ҙࢥܾఆ ࢧԉ ৴པͰ͖Δ ৘ใݯ ݕࡧٕज़

Slide 7

Slide 7 text

Elasticsearch in Our Search

Slide 8

Slide 8 text

ݕࡧΤϯδϯͷྺ࢙ 1999 2007 2014 2023 May 2024 2018 6 8 7 ΦϯϓϨϛε PanaSearch

Slide 9

Slide 9 text

ΞʔΩςΫνϟ ݕࡧج൫ (CPF) ETL APIs DB 12ͷ৘ใαʔϏε 200 ࣾҎ্͔Βίϯςϯπఏܞ ɾ ɾ ɾ + ɾ ɾ ɾ

Slide 10

Slide 10 text

Elasticsearchͷن໛ײ ● ࠷େ 48ϊʔυΫϥελ ● 2ԯ υΩϡϝϯτ ● +1.5ສ / day ● ੑೳ໨ඪ ○ 200 QPS, ≤200 ms Coordinating Only Nodes (3) Data Nodes (42) Master Nodes (3) ɾ ɾ ɾ

Slide 11

Slide 11 text

೔ຊ࠷େڃͷElasticsearchಋೖࣄྫ https://www.elastic.co/jp/customers/nikkei-1 https://www.acroquest.co.jp/archives/15788/

Slide 12

Slide 12 text

࠷ۙͷऔΓ૊Έ: ੜ੒AIؔ࿈

Slide 13

Slide 13 text

3݄17೔ʹRAGαʔϏεΛϦϦʔε༧ఆ

Slide 14

Slide 14 text

No content

Slide 15

Slide 15 text

No content

Slide 16

Slide 16 text

No content

Slide 17

Slide 17 text

No content

Slide 18

Slide 18 text

ݱঢ়ͷΞʔΩςΫνϟ ࣭໰ αϒΫΤϦ ݕࡧ݁Ռ هࣄ ճ౴ Gemini Flash text-embedding -3-large Top 30 results Fetches 1000 docs (200 *5) Gemini Pro knn mseach text- embedding-3- large νϟϯΫ

Slide 19

Slide 19 text

՝୊ײ ● ϋΠϒϦουݕࡧͷಋೖ ○ ݱঢ়ɺϕΫτϧݕࡧͷΈํ͕ਫ਼౓͕ग़͍ͯΔঢ়گ ● ϕΫτϧ༰ྔͷංେԽ ○ 2ԯυΩϡϝϯτΛີϕΫτϧ(1536࣍ݩ)ͰϕΫτϧԽ͢Δͱ3TBҎ্ඞཁ (2.35ഒ) ○ ೔ຊޠεύʔεϕΫτϧϞσϧ ■ ElasticͷELSER͸ӳޠʹ࠷దԽ ■ ೔ຊޠSPLADEϞσϧ͕ొ৔ ○ BBQ (Better Binary Quantization) ■ ES 8.17͔Βಋೖ͞ΕͨϕΫτϧѹॖ

Slide 20

Slide 20 text

·ͱΊ ● ೔ܦͷσʔλϕʔεࣄۀ͸200ࣾҎ্͔Βίϯςϯπఏܞɻ ͦͷ๲େͳίϯςϯπΛԣஅݕࡧͰ͖ΔαʔϏεΛల։ ● ݕࡧ͸զʑͷࣄۀͷίΞٕज़Ͱ͋Γɺͦͷݕࡧ͸ ElasticsearchͰࢧ͑ΒΕ͍ͯΔ ● ੜ੒AIରԠΛੵۃతʹਐΊΔ͜ͱͰɺސ٬ͷҙࢥܾఆʹ͞Β ʹߩݙ

Slide 21

Slide 21 text

We are hiring! https://herp.careers/v1/nikkei/4cJ-o1UA-bz7