search discovery
play

Search + Discovery Peter Bourgon Evolution of search Relevance - PowerPoint PPT Presentation

Search + Discovery Peter Bourgon Evolution of search Relevance ranking A bit about SOA EVOLUTION OF SEARCH KATAMARI DAMACY, NAMCO HOW TO FIX? HOW TO FIX? ElasticSearch HOW TO FIX? ElasticSearch Greenfield development HOW TO FIX?


  1. Search + Discovery Peter Bourgon

  2. Evolution of search Relevance ranking A bit about SOA

  3. EVOLUTION OF SEARCH

  4. KATAMARI DAMACY, NAMCO

  5. HOW TO FIX?

  6. HOW TO FIX? ElasticSearch

  7. HOW TO FIX? ElasticSearch Greenfield development

  8. HOW TO FIX? ElasticSearch Greenfield development Parallel dark launch

  9. HOW TO FIX? ElasticSearch Greenfield development Parallel dark launch Migrate traffic slowly

  10. HOW TO FIX? ElasticSearch Greenfield development Parallel dark launch Migrate traffic slowly Scale out

  11. WHY ELASTICSEARCH

  12. WHY ELASTICSEARCH Good API

  13. WHY ELASTICSEARCH Good API Clear path for growth

  14. WHY ELASTICSEARCH Good API Clear path for growth Batteries included

  15. WHY ELASTICSEARCH Good API Clear path for growth Batteries included Works like you expect it to

  16. COMPONENTS

  17. RELEVANCE RANKING

  18. PageRank

  19. DiscoRank

  20. DISCORANK DETAILS

  21. DISCORANK DETAILS Calculations done offline

  22. DISCORANK DETAILS Calculations done offline Serialize to compact array

  23. DISCORANK DETAILS Calculations done offline Serialize to compact array Load in ES custom scorer

  24. DISCORANK DETAILS Calculations done offline Serialize to compact array Load in ES custom scorer “Hybrid” 2-phase queries

  25. DISCORANK DETAILS Calculations done offline Serialize to compact array Load in ES custom scorer “Hybrid” 2-phase queries Recalculate/reload every N

  26. DISCORANK++

  27. DISCORANK++ + Time component → “Hotness”

  28. DISCORANK++ + Time component → “Hotness” + User behavior → “Trust”

  29. THE BIGGER PICTURE

  30. INDEXING Single-purpose application(s) Stateless and idempotent Build, iterate, iterate, iterate Full catalog ~= 1 hour

  31. SEARCHING Single-purpose application Stateless and idempotent SC/ES DSL translation Open-source ES library

  32. EXPLORING Single-purpose application “Stateless” and idempotent Special, complex searches Cached on regular interval

  33. SERVICE ORIENTED ARCHITECTURE

  34. SOA AT SOUNDCLOUD

  35. SOA AT SOUNDCLOUD Bazooka platform

  36. SOA AT SOUNDCLOUD Bazooka platform API as firewall

  37. SOA AT SOUNDCLOUD Bazooka platform API as firewall 12 Factor applications

  38. 12 FACTOR APPS

  39. 12 FACTOR APPS Single codebase (repository)

  40. 12 FACTOR APPS Single codebase (repository) Config stored in environment

  41. 12 FACTOR APPS Single codebase (repository) Config stored in environment Backing services as resources

  42. 12 FACTOR APPS Single codebase (repository) Config stored in environment Backing services as resources Stateless

  43. 12 FACTOR APPS Single codebase (repository) Config stored in environment Backing services as resources Stateless Horizontal scaling with processes

  44. WHY SOA?

  45. WHY SOA? Velocity

  46. WHY SOA? Velocity Developer happiness

  47. WHY SOA? Velocity Developer happiness Distributed systems are complex

  48. ElasticSearch core + SOA infrastructure

  49. THANKS! (ALSO WE ARE HIRING) Peter Bourgon @peterbourgon http://peter.bourgon.org

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend


More recommend