Replication vs sharding: what is the difference?

Definition:

Replication keeps copies of data so a system can recover from node failure, while sharding distributes different subsets of data across nodes. Replica sets improve redundancy; sharding spreads storage and workload. A sharded deployment can also replicate each shard, so these techniques are complementary.

Key points:

• Replication keeps copies for availability and failover.

• Sharding splits data across servers to spread storage and traffic.

• A poor shard key can send most writes to one shard or require queries across all shards.

• A shard key that increases monotonically can concentrate new writes, while queries without useful shard-key filters may contact many shards.

Interview point:

• Describe distribution, access patterns and hot spots before saying a system can simply “add shards.”