How I Learned to Stop Worrying and Love the Document Store
1 like300 views
The document discusses using MongoDB as a document store for building a landing page hosting platform. It outlines some of the limitations of using SQL and ActiveRecord for this use case, and how MongoDB's flexible document model and horizontal scaling capabilities make it a better fit. Key features of MongoDB that are highlighted include its support for dynamic schemas, embedded collections, and easy aggregation.
How I Learned to Stop Worrying and Love the Document Store
- 1. HOW I LEARNED TO STOP WORRYING AND LOVE THE DOCUMENT STORE COREY EHMKE
- 2. IN A PREVIOUS LIFE... BUILDING A PPC LANDING PAGE HOSTING PLATFORM INTERACTION-LEVEL ANALYTICS A/B AND MULTIVARIATE TESTING DYNAMIC PAGE CONTENT REQUIREMENT FOR REAL-TIME REPORTING
- 3. DOES THIS SOUND FAMILIAR? MODELS CONSTANTLY EVOLVING IN RESPONSE TO ILL- DEFINED REQUIREMENTS AND CHANGING NEEDS POLYMORPHIC REPRESENTATIONS OF USER INTERACTIONS VERSIONING TO SUPPORT MULTIVARIATE TESTING REQUIREMENTS NEED TO QUICKLY SCALE UP IN RESPONSE TO DEMAND
- 4. ACTIVE RECORD: THE WRONG HAMMER BIG DATA = PAINFUL MIGRATIONS RAKE TASKS TO MANAGE DATA TRANSFORMATIONS SINGLE TABLE INHERITANCE NIGHTMARE CHALLENGE OF SCALING UP A SQL DATABASE
- 5. JAMIE ZAWINSKI YOU NEVER KNOW WHAT THE DESIGN IS UNTIL THE PROGRAM IS DONE.
- 6. IN SHORT... ACTIVE RECORD AND SQL JUST DONT FEEL AGILE ENOUGH.
- 7. MONGODB TO THE RESCUE DOCUMENT MODEL ALLOWS FOR THE EASY ADDITION OF NEW FIELDS EMBEDDED COLLECTIONS BUILT-IN VERSIONING MAP REDUCE FOR FAST DATA AGGREGATION SCALING IS EASY
- 8. THE DESIGN PHILOSOPHY DATA MODELS SHOULD BE EASY TO CODE, EASY TO MANAGE, AND HIGHLY PERFORMANT DATABASES SHOULD BE ABLE TO SCALE HORIZONTALLY WITH EASE
- 9. STRENGTHS FLEXIBILITY JSON-LIKE DOCUMENT STORAGE SUPPORT FOR NATIVE TYPES SCHEMA-LESS
- 10. STRENGTHS POWER DYNAMIC QUERIES UPSERTS EASY AGGREGATION VIA MAP-REDUCE
- 11. STRENGTHS SPEED NO JOINS! AUTO-SHARDING TO INCREASE CAPACITY WITHOUT DOWNTIME
- 12. WEAKNESSES LACK OF MATURE GUI TOOLS NO FINE-GRAINED TUNING NO MULTI-DOCUMENT TRANSACTIONS
- 13. PRACTICAL CONSIDERATIONS STOP THINKING IN TABLES, START THINKING IN DOCUMENTS EMBEDDED COLLECTIONS, BUT USE CAREFULLY FULLY CAPABLE OF TRADITIONAL RELATIONS NO SCHEMAS, NO MIGRATIONS NO SQL (USE JAVASCRIPT)
- 14. MONGOID MY OBJECT-DOCUMENT-MAPPER (ODM) OF CHOICE ACTIVEMODEL MAKES MONGOID DOCUMENTS FIRST-CLASS CITIZENS IN RAILS PARTIAL UPDATES (CHANGED FIELDS ONLY) CHAINABLE, LAZY-EVALUATED QUERIES VIA CRITERIA
- 15. MONGODB AT TRUNK CLUB
- 16. NO SCHEMA, NO MIGRATIONS
- 17. NICE QUERY YOU HAVE THERE
- 20. DOES THIS LOOK FAMILIAR?
- 21. HELL YEAH PERFORMANCE 1,000,000 DOCUMENTS 2.93 GHZ IMAC WITH 8GB RAM MODEL#CREATE: 2,392 / SECOND MODEL#UPDATE_ATTRIBUTE: 2,945 / SECOND MODEL.ALL#EACH: 2,325 / SECOND
- 22. EVERYTHING JUST WORKS SCOPES CALLBACKS REFERENCED RELATIONS OBSERVERS VALIDATIONS INDEXES
- 23. USE CASES BIG DATA (LOGS, ANYONE?) COMPLEX QUERIES (MAP REDUCE IS YOUR FRIEND) RAPIDLY EVOLVING MODELS LIBERAL USE OF JSON RELATIONS THAT WANT TO BE EMBEDDED (E.G. ORDERS <=> LINE ITEMS)
- 24. DONT USE IT IF... YOURE WORKING WITH COMPLEX TRANSACTIONS YOURE WORKING WITH MISSION-CRITICAL DATA YOU ENJOY WRITING RAW SQL YOU NEED TO QUICKLY PULL DATA OUT INTO A CSV YOURE AFRAID OF NEW THINGS