Replication in Key-Value Stores
You'll learn how key-value stores copy data across multiple nodes to prevent data loss and keep the system running even when hardware fails.
Why Replicate?
In a key-value store, each piece of data (a key-value pair) lives on a server node. But what happens if that node crashes or the disk fails? The data would be lost. Replication solves this by maintaining multiple copies of each key-value pair across different nodes.
How Replication Works
When you execute a PUT operation to store user:123 → {"name": "Alice"}, the system doesn't just write to one node. Instead:
- Primary node receives the write
- The system forwards copies to two or more replica nodes
- Each replica stores an identical copy
Now if one node fails, the other replicas can still serve GET requests for user:123.
Consistent Hashing and Replicas
Remember consistent hashing from the previous lesson? It determines which node owns a key. With replication, the system typically stores copies on the next N nodes clockwise on the hash ring. For example, with replication factor 3, if user:123 hashes to Node A, copies also go to Nodes B and C.
Durability and Availability
- Durability: Even if a node dies permanently, your data survives on replicas
- Availability: If the primary node is down, replicas can handle
GETandPUToperations
This is crucial for production systems where hardware failures are inevitable.
Key Takeaway: Replication in key-value stores creates multiple copies of each key-value pair across different nodes, ensuring data survives hardware failures and remains accessible even when nodes go down.