Which approach best supports scalability when designing a large data model?

• A. Vertical scaling by hardware upgrade only.
• B. No partitioning and no indexes.
• C. Horizontal partitioning by date or key, indexing on partitioned columns, and ensuring cross-partition query efficiency.
• D. Always keeping all data in a single shard.

Answer: (C)

Scaling a large data model effectively is about spreading data across multiple storage units and keeping queries fast across those units. Horizontal partitioning, such as by date or by a key, distributes rows into separate partitions so work can be done in parallel and data growth doesn’t overwhelm a single node. Putting an index on the partitioned columns helps the system quickly identify which partitions contain the data you need, and designing queries to run efficiently across partitions—with pruning when possible—keeps response times steady as data scales.

Relying on hardware upgrades alone tends to improve capacity temporarily but doesn’t address data growth or distribution. Skipping partitioning and indexing means queries often scan entire datasets, which becomes slow as volume increases. Storing everything in a single shard creates a bottleneck, limiting parallel processing and fault tolerance.