Course contentsShow
← System Design
Lesson 68 of 1,9192. Scaling FoundationsFree lesson

What is Data Replication?

Understanding how keeping copies of data across multiple nodes improves availability and fault tolerance.

What is Data Replication?

What you'll learn: How keeping multiple copies of your data across different machines makes your system more reliable and available.

The Core Idea

Data replication means storing the same data on multiple servers (nodes) instead of keeping it in just one place. When you replicate data, you create identical copies that live on different machines, often in different locations.

Think of it like keeping photocopies of an important document in different safe deposit boxes across town. If one bank burns down, you still have your document safely stored elsewhere.

Why Replication Matters

Without replication, a single server holding your data is a single point of failure. If that machine crashes, catches fire, or loses power, your data becomes unavailable—or worse, lost forever.

With replication:

  • Availability improves: If one server goes down, others can still serve requests
  • Fault tolerance increases: You can survive hardware failures without data loss
  • Read performance scales: Multiple copies mean more servers can handle read requests simultaneously

How It Differs from Partitioning

Remember that partitioning (or sharding) splits your data into separate pieces across nodes—each node holds different data. Replication makes copies of the same data across nodes. These are complementary strategies: you often partition your data AND replicate each partition for redundancy.

Real-World Example

A social media platform might store user profiles. With replication, User #123's profile exists on Server A, Server B, and Server C. If Server A crashes, the application seamlessly reads from Server B or C instead.

Key Takeaway: Data replication creates redundant copies of your data across multiple nodes, ensuring your system stays available and fault-tolerant even when individual machines fail.