Course contentsShow
← System Design
Lesson 200 of 1,9196. Database Scaling: Read ReplicasFree lesson

Why Replicate: Read Scaling

Using read replicas to distribute read load across multiple database instances and improve query throughput.

Why Replicate: Read Scaling

What You'll Learn

You'll discover how database read replicas let you handle more read queries by spreading them across multiple database copies.

The Problem: Read-Heavy Traffic

Most applications perform far more reads than writes—think social media feeds, product catalogs, or news sites. When all read queries hit a single primary database, you eventually exhaust its CPU, memory, and I/O capacity. Response times slow down, and your application can't serve more users.

The Solution: Distribute Read Load

Read replicas create copies of your primary database that handle read-only queries. Instead of one database answering all questions, you now have multiple databases sharing the work.

Think of it like a library: instead of one librarian answering everyone's questions (creating long lines), you hire additional librarians. Each has the same books and can answer questions simultaneously. More librarians = more questions answered per minute.

How It Works

Your application splits traffic by query type:

  • Writes (INSERT, UPDATE, DELETE) → Primary database only
  • Reads (SELECT queries) → Distributed across replicas

With three replicas, you theoretically triple your read throughput. Each replica handles roughly one-third of read queries, reducing load on any single database instance.

Real-World Impact

A single database might handle 1,000 queries/second before slowing down. Add two read replicas, and you can now handle 3,000 queries/second. This horizontal scaling approach grows with your traffic—need more capacity? Add more replicas.

Key Takeaway: Read replicas solve the read scalability bottleneck by distributing SELECT queries across multiple database copies, multiplying your system's total query throughput without requiring a larger primary database.