Course contentsShow
← System Design
Lesson 235 of 1,9197. Database Scaling: ShardingFree lesson

Logical vs Physical Shards

Separating the concept of data partitions from physical servers to enable flexible rebalancing.

Logical vs Physical Shards

What you'll learn: How separating logical data partitions from physical servers gives you flexibility to rebalance and scale your sharded database.


The Problem with Tightly Coupled Shards

When you first implement sharding, you might assign each shard directly to a physical server—Shard 1 lives on Server A, Shard 2 on Server B, and so on. This creates a tight coupling between your data partitions and hardware, making it painful to add servers, move data, or handle uneven growth.

Logical Shards: The Abstraction Layer

Logical shards are data partitions defined by ranges or hash buckets of your shard key—they exist as a concept independent of physical hardware. Think of them as "boxes of data" that can move between servers.

Physical shards (or nodes) are the actual database servers that host those logical shards.

The Key Insight

By creating many more logical shards than physical servers, you gain flexibility. For example, you might have 1,024 logical shards distributed across just 4 physical servers (256 logical shards per server).

Why This Matters

When a physical server becomes overloaded or you add new hardware, you can move individual logical shards between servers without redesigning your entire partitioning scheme. It's like having many small moving boxes instead of one giant immovable container per server.

If Server A is hot and Server D is underutilized, move a few logical shards from A to D. Your application continues using the same shard key logic—only the mapping layer changes.

Key Takeaway: Logical shards decouple your partitioning strategy from physical infrastructure, enabling seamless rebalancing, scaling, and hardware changes without rethinking your shard key or data distribution model.