A MongoDB use case at Telefonica Digital

A success story migrating from Oracle to MongoDB. The same than https://speakerdeck.com/pablito56/from-oracle-to-mongodb and https://speakerdeck.com/pablito56/from-oracle-to-mongodb-real-uses-cases-fib-upc but more business oriented and less technical details.

Transcript

1. MongoDB NYC 2013 A MongoDB use case at Telefonica Digital Pablo Enfedaque @pablitoev56
2. 2 Telefonica Digital Executive summary •  We built the Personalisation Server using Oracle 11g •  We had performance issues with Oracle •  We built a new version with MongoDB in 4 months half the team •  We obtained great benefits §  Performance boost (one order of magnitude), predictable scaling §  Low time to market and better extensibility §  New opportunities for other products and services 01
3. Título del capítulo Máximo 3 líneas 01 Telefonica Digital. Who we are?
4. 4 Telefonica Digital Who we are •  Telefonica §  Fourth largest telecommunications company in the world §  Operations in Europe (7 countries), USA and Latam (15 countries) §  315M of customers with Movistar, O2, Vivo, Terra… •  Telefonica Digital §  “Beyond connectivity”, division for web and mobile services & contents §  Jajah, Tokbox, Terra, TU, OWD / Firefox OS… 01
5. 5 Telefonica Digital Who we are •  Telefonica Product Development and Innovation unit •  Around 70 different on going projects §  1 year, 6 months, 3 months or even 10 days projects 01
6. Título del capítulo Máximo 3 líneas 02 Personalisation Server. Why?
7. 7 Telefonica PDI Personalisation Server. Why? •  We were storing profile data of millions of customers. However… 01
8. 8 Telefonica PDI Personalisation Server. Why? •  We were storing profile data of millions of customers. However… §  Each service had its own internal storage §  Customers profile was scattered across a variety of different DBs §  In some cases stored in a DWH, others in isolated DBs §  No homogeneous interfaces or data structure §  Repeated or out-dated information 01
9. 9 Telefonica PDI Personalisation Server. Why? •  We were storing profile data of millions of customers. However… Customers data was neither shared nor usable 01
10. 10 Telefonica PDI Personalisation Server. Why? 01
11. 11 Telefonica PDI Personalisation Server. Why? •  Personalisation Server as master customer’s data storage §  Operational storage, real-time access §  Flexible customer’s profile structure, classified in services §  Data sharing across all services §  Massive off-line batch interface §  Authentication and fine grained authorization §  Fault tolerance and high availability §  Inexpensive solution, low hardware requirements 01
12. Título del capítulo Máximo 3 líneas 03 Oracle-based solution
13. 13 Telefonica PDI Oracle-based solution •  20 people involved (7 developers) in 14 – 15 months •  Oracle 11g Standard or Enterprise •  Three instances deployed •  Three iterations, focusing on performance enhancements 01
14. 14 Telefonica PDI After three iterations •  Three successful deployments, but… §  Oversized storage requirements §  Performance below expectations Still not scaling enough 04
15. 15 Telefonica PDI The fourth deployment arrived •  Everything went fine until one month before deploying •  New performance requirements appeared §  40% more data to be stored and processed §  Reload the whole DWH (22M customers) daily in small time window •  Servers already purchased 01
16. 16 Telefonica PDI Huge improvement needed •  Reduce storage to a half •  Enhance performance up to 3 times •  Only 4 months with a reduced team Fourth iteration with Oracle or move to a new technology? 01
17. Título del capítulo Máximo 3 líneas 04 MongoDB-based solution
18. 18 Telefonica PDI Why MongoDB •  Seemed to be mature enough •  Several features matching our requirements §  Dynamic schema §  Fault tolerance and high availability §  Official drivers §  Production support §  Reduced costs •  We did some tests and performed very well 06
19. 19 Telefonica Digital MongoDB-based solution •  We built a new version of the PS with MongoDB 2.0 •  Half the team (3 – 4 developers) •  One fourth of the time (4 months) •  We even enhanced some features or added new ones 01
20. 20 Telefonica Digital MongoDB performance •  Performance boost of more than one order of magnitude •  Lower storage requirements (one third) •  Simple, extensible and maintainable schema 01 MONGODB ORACLE
21. 21 Telefonica PDI MongoDB-based solution 01
22. Título del capítulo Máximo 3 líneas 05 Conclusions
23. 23 Telefonica PDI Lessons learned •  Awesome performance boost with MongoDB •  Really fast and agile development •  New technology, understand how to use it •  Beat the fear of the unknown. It was mature enough for us •  Great ecosystem & tools, great support •  Lots of new solutions and use cases 06
24. 24 Telefonica PDI Results for Telefonica •  Better products performance •  Faster and easier development •  Increased customer satisfaction •  Decreased costs, increased revenue •  Opportunities for new products and services §  Firefox OS push notifications §  Telefonica Machine-to-machine solutions 06
27. Q & A Thanks!
28. How MongoDB Helps Telefonica Digital, Pablo Enfedaque Next Sessions at 2:50 5th Floor: West Side Ballroom 3&4: Data Modeling Examples from the Real World West Side Ballroom 1&2: Growing Up MongoDB Juilliard Complex: Business Track: MetLife Leapfrogs Insurance Industry with MongoDB-Powered Big Data Application Lyceum Complex: Ask the Experts: MongoDB Monitoring and Backup Service Session 7th Floor: Empire Complex: How We Fixed Our MongoDB Problems SoHo Complex: High Performance, High Scale MongoDB on AWS: A Hands On Guide