비정규화 전략

In traditional relational databases, we "normalize" data to avoid redundancy. But NoSQL databases like Firebase Realtime Database often benefit from the opposite: denormalization.

Denormalization involves intentionally adding redundant data or grouping data to optimize read performance. It's a key strategy for speed in NoSQL.

Firebase Realtime Database excels at real-time updates and quick reads. However, complex queries across multiple data paths can be slow and costly.

• Flat Structures: Firebase works best with flat data structures.
• Reduced Reads: Denormalization can reduce the number of reads needed for common queries.
• Query Limitations: NoSQL databases have limited querying capabilities compared to SQL.

There are several ways to denormalize your data. Two common patterns are:

• Duplication: Storing the same piece of data in multiple locations.
• Aggregation: Storing pre-calculated summaries or counts.
• Fan-out: Writing data to multiple paths simultaneously (often used with duplication).

We'll focus on the first two in detail.

Imagine a social app where users post messages. Each post needs to display the author's name.

Instead of fetching the user's profile every time you display a post, you can duplicate the user's name directly into the post object when it's created. This makes displaying posts much faster.

// Original structure (normalized)
posts: {
  postId1: {
    text: "My first post!",
    authorId: "userId123"
  }
},
users: {
  userId123: {
    name: "Alice"
  }
}

// Denormalized structure
posts: {
  postId1: {
    text: "My first post!",
    authorId: "userId123",
    authorName: "Alice" // Duplicated!
  }
}

Here's how you might write data to include duplicated user information. When a user posts, we add their name directly to the post object.

Another common denormalization technique is to store aggregated data, like counts or sums, directly on a parent object.

For instance, if you have a blog post and want to quickly show the number of comments, you can store a commentCount property on the post itself. This avoids fetching all comments just to count them.

// Without aggregation
posts: {
  postId1: {
    title: "My Blog Post"
  }
},
comments: {
  commentId1: { postId: "postId1", text: "..." },
  commentId2: { postId: "postId1", text: "..." }
}

// With aggregation
posts: {
  postId1: {
    title: "My Blog Post",
    commentCount: 2 // Aggregated!
  }
}

When a new comment is added, we can increment a counter on the parent post. This ensures the count is always up-to-date and easily accessible.

Denormalization is a powerful tool, but it's not always the answer. Consider it when:

• Read Performance is Critical: You need to display data quickly and frequently.
• Queries are Complex: Your app often needs to fetch related data that lives in different paths.
• Data Changes Infrequently: The duplicated or aggregated data doesn't change often.

Always weigh the benefits against the complexity.

The main challenge with denormalization is maintaining data consistency. If you duplicate a user's name, and they change it, you need to update it in all duplicated locations.

• Increased Write Complexity: More writes are needed to keep data in sync.
• Potential for Inconsistency: If an update fails, data can become out of sync.
• Cloud Functions: Firebase Cloud Functions are often used to automate consistency for denormalized data.

We've learned that denormalization is a key technique for optimizing read performance in Firebase Realtime Database by intentionally adding redundant or aggregated data.

• It helps create flatter data structures.
• Common patterns include duplication and aggregation.
• It speeds up reads but adds complexity to writes and requires careful consistency management, often with Cloud Functions.

Consider your most frequent read patterns when deciding where and how to denormalize.