MongoDB Data Modeling

Steps to take

1. Identify workload

2. Map relationship

3. Apply design patterns

4. Optimization - creating indexes

Identify workload

At this step, we mostly need to identify if our application is read or write-heavy and which resources are critical.

Example

Data relationship

Referencing or Embedding

Example of referencing

// Referencing
// books
{
  "_id": "book1",
  "title": "MongoDB",
  "authors": ["author1", "author2"]
}

// authors
[
  { "author_id": "author1", "name": "Huy" },
  { "author_id": "author2", "name": "Steve" }
]

Example of embedding

{
  "_id": "book1",
  "title": "MongoDB",
  "authors": [
    { "author_id": "author1", "name": "Huy" },
    { "author_id": "author2", "name": "Steve" }
  ]
}

When to use which?

In MongoDB, the recommended way to handle related data is to embed it into a sub-document - retrieve information without $lookup. The other way is to use references.

👉 The golden rule: Data that is accessed together should be stored together

Let’s take a look at the above books and authors relationship

Here are the guidelines

• Simplicity - would keeping the pieces together lead to a simpler data model and code? => Yes - Embed

• Go Together - Do the pieces of information have a “has-a”, “contains” relationship? Yes - Embed

• Query Automicity - Does the app query the pieces of info together? Yes - Embed

• Update Complexity - Are the pieces of info updated together? No - Reference

  • We can update the book and author separately

• Archival - should the pieces of information be archived at the same time? No - Reference

• Cardinality * - is there a high growth in the child side of the relationship? No - Embed

• Data Duplication - Would data duplication be too complicated to manage and undesired? No - Embed

• Document Size - Would the combined size take too much memory or transfer bandwidth? No - Embed

• Document Growth - would the embedded piece grow without bound? No - Embed

• Workload - are the pieces of info written at different times in a write-heavy workload? No - Embed

  • If we frequently update related data, then better to use a reference

• Individuality - for the child side, can the pieces exist by themselves without a parent? No - Embed

  • An author cannot exist in our app without a book

👉 But mostly consider

• Array size

• Update frequency of embedded data

• Whether the embedded entity is often queried independently

One-to-one

Example: one company has one headquarters

Embedded - recommended in this case ✅

{
  "_id": "publisher123",
  "name": "MongoDB Press",
  "street": "123 Data Lane",
  "city": "San Francisco",
  "state": "CA",
  "country": "USA",
  "zip": "94107"
}

// it's better to group - nested object
{
  "_id": "publisher123",
  "name": "MongoDB Press",
  "headquarters": {
    "street": "123 Data Lane",
    "city": "San Francisco",
    "state": "CA",
    "country": "USA",
    "zip": "94107"
  }
}

Reference

Option 1: Array of references within the parent document

✅ Use when - the app primarily reads publisher documents

// publishers
{
  "_id": "publisher123",
  "name": "MongoDB Press",
  "headquarter_id": "hq1"
}

// headquarters
{
  "_id": "hq1",
  "street": "123 Data Lane",
  "city": "San Francisco",
  "state": "CA",
  "country": "USA",
  "zip": "94107"
}

Option 2: a child document to reference the parent document

✅ Use when - the app primarily reads headquarters documents

// publishers
{
  "_id": "publisher123",
  "name": "MongoDB Press"
}

// headquarters
{
  "_id": "hq1",
  "publisher_id": "publisher123",
  "street": "123 Data Lane",
  "city": "San Francisco",
  "state": "CA",
  "country": "USA",
  "zip": "94107"
}

One-to-Many

Example: one printed book has many reviews

Embedded

Good when

• No duplication of information - i.e, in this case, review can be attached to one book only

• The option on many sides cannot exist by itself - i.e, review cannot exist without the printed book

Bad when

• Children are an unbounded array => it can make the document super large

Option 1: As an array

{
  "_id": "book0456",
  "title": "Practical MongoDB Aggregations",
  "author": "Paul Done",
  "reviews": [
    {
      "user_id": "user0956",
      "title": "Lorem Ipsum",
      "review_text": "dolor sit amet...",
      "rating": 3
    },
    {
      "user_id": "user0345",
      "title": "Ut enim",
      "review_text": "ad minim veniam...",
      "rating": 4
    }
  ]
}

Option 2: As an object

{
  "_id": "book0456",
  "title": "Practical MongoDB Aggregations",
  "author": "Paul Done",
  "reviews": {
    "user0956": {
      "title": "Lorem Ipsum",
      "review_text": "dolor sit amet...",
      "rating": 3
    },
    "user0345": {
      "title": "Ut enim",
      "review_text": "ad minim veniam...",
      "rating": 4
    }
  }
}

Reference

Option 1: Array of references to child documents in parent document

Use when: The list of children is small and often accessed together with the parent

// books
{
  "_id": "book0456",
  "title": "Practical MongoDB Aggregations",
  "author": "Paul Done",
  "reviews": [
    "review001",
    "review002"
  ]
}

// reviews
[
  {
    "_id": "review001",
    "user_id": "user0956",
    "title": "Lorem Ipsum",
    "review_text": "dolor sit amet...",
    "rating": 3
  },
  {
    "_id": "review002",
    "user_id": "user0345",
    "title": "Ut enim",
    "review_text": "ad minim veniam...",
    "rating": 4
  }
]

Option 2: a child document to reference the parent document - recommended in this case due to an unbounded array ✅

Use when:

• The number of children is large or grows unbounded.

• You rarely need all the children at once.

// books
{
  "_id": "book0456",
  "title": "Practical MongoDB Aggregations",
  "author": "Paul Done"
}

// reviews
[
  {
    "_id": "review001",
    "book_id": "book0456",
    "user_id": "user0956",
    "title": "Lorem Ipsum",
    "review_text": "dolor sit amet...",
    "rating": 3
  },
  {
    "_id": "review002",
    "book_id": "book0456",
    "user_id": "user0345",
    "title": "Ut enim",
    "review_text": "ad minim veniam...",
    "rating": 4
  }
]

Many-to-many

Example: a book has many authors, and an author can write many books

Embedded - recommended in this case, check the “When to use which” section above ✅

Notice that here we have data duplication. However, it’s not always an issue

Option 1: As an array

{
  "_id": "book0008",
  "title": "Mastering MongoDB 7.0",
  "authors": [
    {
      "author_id": "author0021",
      "name": "Elie Hannouch"
    },
    {
      "author_id": "author0087",
      "name": "Malak Abu Hammad"
    },
    {
      👉 "author_id": "author0046",
      "name": "Rachelle Palmer"
    }
  ]
}

// another book
{
  "_id": "book0017",
  "title": "Building Rich and Resilient Applications with MongoDB",
  "authors": [
    {
      👉 "author_id": "author0046",
      "name": "Rachelle Palmer"
    }
  ]
}

Option 2: As an object

{
  "_id": "book0008",
  "title": "Mastering MongoDB 7.0",
  "authors": {
    "author0021": {
      "name": "Elie Hannouch"
    },
    "author0087": {
      "name": "Malak Abu Hammad"
    },
    "author0046": {
      "name": "Rachelle Palmer"
    }
  }
}

Reference

Option 1: Array of references to child documents in parent document

// books
[
  {
    "_id": "book0008",
    "title": "Mastering MongoDB 7.0",
    "authors": [
      "author0021",
      "author0087",
      👉 "author0046"
    ]
  },
  {
    "_id": "book0017",
    "title": "Building Rich and Resilient Applications with MongoDB",
    "authors": [
      👉 "author0046"
    ]
  }
]

// authors
[
  {
    "_id": "author0021",
    "name": "Elie Hannouch"
  },
  {
    "_id": "author0087",
    "name": "Malak Abu Hammad"
  },
  {
    "_id": "author0046",
    "name": "Rachelle Palmer"
  }
]

Option 2: a child document to reference the parent document

// books
[
  {
    "_id": "book0008",
    "title": "Mastering MongoDB 7.0"
  },
  {
    "_id": "book0017",
    "title": "Building Rich and Resilient Applications with MongoDB"
  }
]

// authors
[
  {
    "author_id": "author0021",
    "name": "Elie Hannouch",
    // not that unlike the one-to-many
    // this is an array
    👉 "books": ["book0008"]
  },
  {
    "author_id": "author0087",
    "name": "Malak Abu Hammad",
    👉 "books": ["book0008"]
  },
  {
    "author_id": "author0046",
    "name": "Rachelle Palmer",
    "books": ["book0008", "book0017"]
  }
]

Resources

• https://www.mongodb.com/docs/manual/data-modeling/

• https://learn.mongodb.com/learning-paths/data-modeling-for-mongodb

• https://www.mongodb.com/docs/manual/data-modeling/concepts/embedding-vs-references/