Which condition most clearly triggers the need to shard a database?

Which condition most clearly triggers the need to shard a database?

• A. Shard when storage exceeds what a single instance can handle.
• B. Shard when read load cannot be addressed by read replicas alone.
• C. Shards when data fits easily on a single node.
• D. Shard when write throughput is well below capacity of a single database.

Answer: (A)

Sharding is used when a single database instance can no longer store all the data or sustain the write workload. The clearest trigger is that the dataset won’t fit on one node, so spreading data across multiple nodes increases total storage and distributes writes across partitions. Read replicas can help with read traffic, but they don’t expand the total storage or write capacity of the dataset. If everything fits on one node, there’s no need to shard, and if the write load is already well within capacity, adding shards would add unnecessary complexity.

Sharding is used when a single database instance can no longer store all the data or sustain the write workload. The clearest trigger is that the dataset won’t fit on one node, so spreading data across multiple nodes increases total storage and distributes writes across partitions. Read replicas can help with read traffic, but they don’t expand the total storage or write capacity of the dataset. If everything fits on one node, there’s no need to shard, and if the write load is already well within capacity, adding shards would add unnecessary complexity.