MongoDB Data Modeling

Han Yi

Know your business

Determine your position

2.Data Model

Logical Model

Physical Model

1.Domain Model

?

Master your kits

vs

Core concept: Data as a Document

Why Document ?

• Humongous support
  • ​DoubleClick, 2007
  • 400000 ads/s
• Flexible types for data model
  • Key-Value
  • List
  • Rich document
  • Embedded array and document
  • ...

Pros vs Cons of document model

• Pros
  • ​Flexible data model design
  • High performance read
• Cons
  • Design heavily depends on scenario
  • Single document should be no larger than 16mb, no deeper than 100 layers

Data model patterns for a document

• Normalized
  • referenced in different document
• Non-normalized
  • nested in single document
• Tree structure
  • hybrid pattern

MongoDB prefer

• Non-normalized

Real case

• Fan Out Read
• Fan Out Write

3 Factors of data modeling

• Performance
• Query pattern
• Life cycle of data and growth rate of documents (Application domain)

Sample 1: Date modeling

• BSON Date type is 64 bits signed number
• BSON Date type is UTC time
• Time zone is suggested using additional fields:

var now = new Date();

db.data.save( { date: now,

                            offset: now.getTimezoneOffset() } );

Sample 2: Monetary data modeling

• Numeric Model
  • Decimal format (3.4+)
  • IEEE 754-2008: Decimal 128
  • Type long (64 bits) + Power factor (older than 3.4)
• Non-numeric Model
  • Full value: String
  • Approximate value: double

Sample 3: Content query modeling

• $regex
• $text
• Keywords array: multikey index