Which statement best captures when to shard?

Which statement best captures when to shard?

• A. Shard when storage or write throughput exceeds what a single database can handle, or read load cannot be addressed by read replicas.
• B. Shard only when data exceeds 10 TB.
• C. Shard as soon as a project starts, regardless of load.
• D. Shard only for analytic processing workloads.

Answer: (A)

Sharding is about scaling a database by distributing data and load across multiple nodes to increase capacity and throughput. The best rule of thumb is to shard when the current single database can’t meet demands for storage or write throughput, or when read traffic is so high that adding read replicas no longer solves the latency or throughput goals. Read replicas can help, but they don’t reduce the number of writes or the amount of data stored; sharding spreads both across multiple writable nodes, which is essential when those limits are reached. This decision isn’t about a fixed data size like a certain terabyte threshold, nor about starting to shard as soon as a project begins. It also isn’t limited to analytic workloads—OLTP systems can need sharding too if they outgrow a single instance. Keep in mind that sharding adds complexity: you’ll need a good shard key, routing logic, and strategies for cross-shard operations and rebalancing, so you typically shard only when scaling requirements justify that extra complexity.

Sharding is about scaling a database by distributing data and load across multiple nodes to increase capacity and throughput. The best rule of thumb is to shard when the current single database can’t meet demands for storage or write throughput, or when read traffic is so high that adding read replicas no longer solves the latency or throughput goals. Read replicas can help, but they don’t reduce the number of writes or the amount of data stored; sharding spreads both across multiple writable nodes, which is essential when those limits are reached.

This decision isn’t about a fixed data size like a certain terabyte threshold, nor about starting to shard as soon as a project begins. It also isn’t limited to analytic workloads—OLTP systems can need sharding too if they outgrow a single instance. Keep in mind that sharding adds complexity: you’ll need a good shard key, routing logic, and strategies for cross-shard operations and rebalancing, so you typically shard only when scaling requirements justify that extra complexity.