Your SlideShare is downloading. ×
Neo4j au coeur du graphe social de 45 millions de membres par Nicolas Tricot
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×

Saving this for later?

Get the SlideShare app to save on your phone or tablet. Read anywhere, anytime - even offline.

Text the download link to your phone

Standard text messaging rates apply

Neo4j au coeur du graphe social de 45 millions de membres par Nicolas Tricot

1,765
views

Published on

Neo4j au coeur du graphe social de 45 millions de membres", ou comment Viadeo est passés d'une technologie maison devenue limitée à un graphe-database plein de perspectives d'avenir pour modéliser son …

Neo4j au coeur du graphe social de 45 millions de membres", ou comment Viadeo est passés d'une technologie maison devenue limitée à un graphe-database plein de perspectives d'avenir pour modéliser son graphe social...
http://fr.viadeo.com/fr/profile/nicolas.tricot


0 Comments
1 Like
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total Views
1,765
On Slideshare
0
From Embeds
0
Number of Embeds
4
Actions
Shares
0
Downloads
7
Comments
0
Likes
1
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide

Transcript

  • 1. Your network is more powerful than you thinkNeo4J au cœur du graphe social de 45 millions de membres Viadeo Tech Days Les 20, 21 et 22 novembre 2012 1
  • 2. ABOUT THE VIADEO GROUP • 1 million new members / month • 10 million connexions / month • 100 million profiles viewed / monthYour network is more powerful than you think 2 / 36
  • 3. GRAPHS ARE EVERYWHERE
  • 4. GRAPHS ARE EVERYWHEREYour network is more powerful than you think 4 / 36
  • 5. GRAPHS ARE EVERYWHEREYour network is more powerful than you think 5 / 36
  • 6. GRAPHS ARE EVERYWHERE © Air France / KLMYour network is more powerful than you think 6 / 36
  • 7. GRAPHS ARE EVERYWHEREYour network is more powerful than you think 7 / 36
  • 8. SOCIAL GRAPH
  • 9. SOCIAL GRAPH Nodes RelationshipsYour network is more powerful than you think 9 / 36
  • 10. SOCIAL GRAPH Direct contactsYour network is more powerful than you think 10 / 36
  • 11. SOCIAL GRAPH Level 2 contactsYour network is more powerful than you think 11 / 36
  • 12. SOCIAL GRAPH Path Distance 3 Path Distance 4Your network is more powerful than you think 12 / 36
  • 13. PREHISTORY2006 - 2011
  • 14. PREHISTORY 2006-2011 • In-house algorithm • Network storage in MySQL Database CREATE TABLE `Network` ( `memberId` int(11) NOT NULL DEFAULT 0, `L1` mediumblob NOT NULL, `L2` mediumblob NOT NULL, PRIMARY KEY (`memberId`) ) ENGINE=InnoDB;Your network is more powerful than you think 14 / 36
  • 15. PREHISTORY 2006-2011 Update the network (old-fashioned style) Member A and Member B are now in contacts Update of A.L1 + B.L1 and A.L2 + B.L2 Retrieving A.L1 + B.L1 and update *.L2 Example: • A has 500 contacts • B has 150 contacts 500 + 150 + 2 = 652 updates!Your network is more powerful than you think 15 / 36
  • 16. PREHISTORY 2006-2011 Good performances on Computation of Paths Computation of DistancesYour network is more powerful than you think 16 / 36
  • 17. PREHISTORY 2006-2011 = =Your network is more powerful than you think 17 / 36
  • 18. PREHISTORY 2006-2011 LIMITATIONS 1) Important latency for complete update 2) Massive bandwidth impact for internal network 3) 48 hours to restart from scratchYour network is more powerful than you think 18 / 36
  • 19. GRAPH DATABASE
  • 20. GRAPH DATABASE Name: Graham Greene Name: Graham Greene Title: Tinker, Soldier, Spy Title: Our man in Havana Born: 02-10-1904 Born: 19-10-1932 Published: 1974 Published: 1958 Died: 02-04-1991 1 WROTE 2 4 WROTE 3 RECOMMENDED Date: 09-09-2011 RECOMMENDED RECOMMENDED Date: 05-07-2011 Date: 03-02-2011 6 5 Name: Alan Name: Ian © Ian RobinsonYour network is more powerful than you think 20 / 36
  • 21. GRAPH DATABASE Name: Graham Greene Name: Graham Greene Title: Tinker, Soldier, Spy Title: Our man in Havana Born: 02-10-1904 Born: 19-10-1932 Published: 1974 Published: 1958 Died: 02-04-1991 1 WROTE 2 4 WROTE 3 RECOMMENDED Date: 09-09-2011 RECOMMENDED RECOMMENDED Date: 05-07-2011 Date: 03-02-2011 Nodes 6 5 Name: Alan Name: Ian © Ian RobinsonYour network is more powerful than you think 21 / 36
  • 22. GRAPH DATABASE Name: Graham Greene Name: Graham Greene Title: Tinker, Soldier, Spy Title: Our man in Havana Born: 02-10-1904 Born: 19-10-1932 Published: 1974 Published: 1958 Died: 02-04-1991 1 WROTE 2 4 WROTE 3 RECOMMENDED Date: 09-09-2011 RECOMMENDED RECOMMENDED Date: 05-07-2011 Date: 03-02-2011 Nodes 6 5 Relationships Name: Alan Name: Ian © Ian RobinsonYour network is more powerful than you think 22 / 36
  • 23. GRAPH DATABASE Name: Graham Greene Name: Graham Greene Title: Tinker, Soldier, Spy Title: Our man in Havana Born: 02-10-1904 Born: 19-10-1932 Published: 1974 Published: 1958 Died: 02-04-1991 1 WROTE 2 4 WROTE 3 RECOMMENDED Date: 09-09-2011 RECOMMENDED RECOMMENDED Date: 05-07-2011 Date: 03-02-2011 Nodes 6 5 Relationships Name: Alan Name: Ian Properties © Ian RobinsonYour network is more powerful than you think 23 / 36
  • 24. WHY NEO4J
  • 25. WHY Neo4J Findings after POC on 3 other tools: • Old technology with add-on for graph management • No user communities • Bad performance • “Black Box” code Why ? • OpenSource project • Good documentation • User community • Excellent performance • ACID • Very simple • (How to better model a Social Graph than with a Graph database ?!?)Your network is more powerful than you think 25 / 36
  • 26. WHY Neo4J 1 node = 1 member 1 Relationship = 1 direct contactYour network is more powerful than you think 26 / 36
  • 27. WHY Neo4J BENEFITS Very easy to integrate (less than 2 months) Instantaneous graph updates High Availability Backup / RestoreYour network is more powerful than you think 27 / 36
  • 28. LIMITATION
  • 29. LIMITATION Doesn’t handle SHARDING! (Split one graph onto several servers) « Size doesn’t matter… », but… Server 1 Server 2Your network is more powerful than you think 29 / 36
  • 30. EXPLORATION MODE
  • 31. EXPLORATION MODE What for the future? Store various kind of objects Change the development paradigmYour network is more powerful than you think 31 / 36
  • 32. EXPLORATION MODEYour network is more powerful than you think 32 / 36
  • 33. EXPLORATION MODEYour network is more powerful than you think 33 / 36
  • 34. CONCLUSION
  • 35. CONCLUSION Neo4J: Has replaced a 5-year-old in-house technology in only 2 months Supports the core system of the Viadeo Professional Social Network Has been in production for 1 year ½ Deals smoothly with Viadeo’s usage growth Think about how Neo4J will improve your own business!Your network is more powerful than you think 35 / 36