Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Big Dating at eHarmony


Published on

eHarmony uses MongoDB to make it easier for couples to find each other. At the forefront of big data and machine learning, eHarmony’s matching system uses a flow algorithm to process a billion potential matches per day. The Compatibility Matching System® uses bi-directional, user-defined criteria to match members based on a comprehensive set of traits and preferences. The system was originally built on a relational database, but with over 51MM+ users, it took more than 2 weeks for the matching algorithm to run. By switching to MongoDB, eHarmony reduced the time to match by 95% to under 12 hours. Learn about eHarmony's matching system, its technology evaluation process, and how it has used MongoDB to make the happiest couples in the world.

Published in: Technology

Big Dating at eHarmony

  1. 1. Thod  Nguyen   Chief  Technology  Officer   Big Dating at eHarmony
  2. 2. social impact
  3. 3. big dating at scale !   3B+  poten9al  matches  daily  ~  25+  TB  of  data   !   60M+  mul9-­‐aDribute  queries  daily  looking  across  250+   aDributes       !   212M+  photos  ~  15+  TB  of  data   !   4B+  rela9onship  ques9onnaires  ~  25+  TB  of  data    
  4. 4. the big win for product  Decreased  the  processing  0me  to  match  by  95%,     from  2+  weeks  to  12  hours   on  3B+  poten0al  matches/day     !   30%  increase  in  2-­‐way  communica9ons   !   50%  increase  in  paid  subs   !   60%  increase  in  unique  visitors        
  5. 5. today     !   Compa9bility  Matching  System   !      The  Old   !      The  New   !      Why  MongoDB   !      What’s  Next    
  6. 6. compatibility matching system®   Compa0bility  Matching  System®   Match   Distribu0on   3! Compa0bility     Matching   1! Affinity     Matching   2!
  7. 7. Compa0bility  Matching  System®   Affinity     Matching   Match   Distribu0on   2! 3! compatibility matching system (cont’d) Compa0bility     Matching   1!
  8. 8. traditional search
  9. 9. eharmony matching
  10. 10. compatibility models
  11. 11. compatibility matching process
  12. 12. legacy compatibility match processor (CMP)
  13. 13. legacy compatibility match processor V.2 (CMP)
  14. 14. challenges with existing v2. design
  15. 15. challenges with existing v2. design (contd.)
  16. 16. challenges with existing v2. design (contd.)
  17. 17. challenges with existing v2. design (contd.)
  18. 18. challenges with existing v2. design (contd.)
  19. 19. new data store requirements
  20. 20. why Mongodb?
  21. 21. tradeoffs !   No  schema  =  larger  footprint   !   Aggrega9on  queries  are  different     !   Ini9al  configura9on  can  be  long,  manual  process    
  22. 22. lessons learned !   Turn  on  the  Firehose   !   Unleash  the  Chaos  Monkey   !   Engage  MongoDB,  Inc.  early  –  dev  to  produc9on   !   Try  to  isolate  your  queries  to  a  shard   !   Run  in  shadow  mode  
  23. 23. what’s next New  matching  use  cases:     !   Globaliza9on  and  Localiza9on  of  eH  site   !   Careers  by  eHarmony   !   Internet  of  Things  “Compa9ble”     New  use  cases  within  eHarmony:     !   Real-­‐9me  geo  loca9on  based  matching  service     !   Careers  
  24. 24. technology stack
  25. 25.   We’re  Hiring