Test SQL With Realistic Data Volumes
A query that is instant on ten rows can fail under production-sized data. This guide gives you a small, testable way to approach the issue before changing a live system.
A practical workflow
- Build a scrubbed or synthetic dataset with similar row counts, distributions and indexes.
- Measure representative reads and writes, then inspect the plan and memory use.
What to check
Never copy customer data into an unprotected test environment; preserve privacy while reproducing scale. Compare the result with the original problem and record what changed. If a measurement or test does not support the change, revisit the assumption rather than adding more complexity.
Keep the experiment small enough to reverse. Document the starting state, the change and the result so another person can repeat the process. Product menus and versions can change, so check the current documentation for your environment.
Official reference: Read the documentation.