Study SnowPro COF-C02 Warehouses and Caching: key concepts, common traps, and exam decision cues.
Warehouse questions are often tradeoff questions. COF-C02 wants you to understand when sizing helps, when concurrency is the real problem, and when cached behavior explains the performance difference without more compute.
| Requirement | Strongest first lens |
|---|---|
| one workload needs more raw compute | warehouse sizing |
| many users compete at once | concurrency behavior |
| repeated similar queries seem faster | caching awareness |
| credit spend is drifting upward | auto-suspend and workload separation |
| Ask this first | Why it matters |
|---|---|
| is the issue one heavy workload or many competing workloads? | that separates sizing from concurrency |
| is the speedup real, or is cache behavior masking the baseline? | repeated runs can mislead |
| is the real problem cost during idle time? | operational warehouse settings may matter more than SQL changes |
| Trap | Better rule |
|---|---|
| treating cached speedup as permanent workload proof | cache effects can hide the real baseline |
| solving concurrency pain with storage explanations | concurrency lives in the compute lane |
| leaving warehouses running unnecessarily | suspend behavior is part of cost control |
Snowflake wants you to notice that “slow,” “busy,” and “expensive” are not the same thing:
The stronger answer identifies which problem you actually have.
| Scenario clue | Stronger answer shape |
|---|---|
| “same query is faster on repeat with no data change” | cache-aware reading |
| “many users compete at once” | concurrency lane |
| “warehouse keeps running after work ends” | auto-suspend and operational control |
| “single workload just needs more compute” | sizing lane |