What Is Database Sharding?

Sharding splits a single logical database's data across multiple physical databases (shards), each holding a subset of the data, to scale beyond what one dat...

Published November 20, 2026

Sharding splits one large database into smaller pieces (shards) spread across servers, each holding a subset of the rows, so data and traffic scale beyond a single machine.

The concept

A shard key (user_id, region) decides which shard stores a row, typically by hashing or ranges. Queries with the key go to one shard; others must fan out.

Sharding adds complexity: cross-shard joins and transactions are hard, rebalancing is painful and a bad shard key creates hot spots.

Route by key

query→ shard 1Appuser_id = 42Routerhash(42) % 3Shard 1holds user 42
The shard key picks the server.

Practical examples

Hash routing

const shards = [db0, db1, db2]
const shardFor = (userId) => shards[userId % shards.length]
await shardFor(42).query('SELECT * FROM orders WHERE user_id = ?', [42])

Range sharding

Shard A: users 1–1M · Shard B: 1M–2M

Before sharding try

Indexes, caching, read replicas, bigger hardware, partitioning

Common mistakes

  • Sharding too early.
  • A shard key that concentrates load (hot shard).
  • Needing cross-shard joins everywhere.

Key takeaways

  • Horizontal split of data.
  • Shard key is crucial.
  • Last resort after simpler scaling.

FAQ

Sharding vs partitioning?

Partitioning splits a table inside one database; sharding spreads across servers.

Replication vs sharding?

Replication copies the same data; sharding divides different data.

More SQL articles