COF-C03 Question 66
Single answerScaling policiesA data engineering team uses a multi-cluster virtual warehouse named ETL_WH to support dozens of concurrent ELT jobs that start at the top of every hour. The warehouse is configured as a multi-cluster warehouse with MIN_CLUSTER_COUNT = 1 and MAX_CLUSTER_COUNT = 4. The team wants to improve query response times during these concurrency spikes, even if it increases compute cost. Which scaling policy should they choose for ETL_WH?
- A
STANDARD, because it favors keeping additional clusters available to handle query bursts sooner
- B
ECONOMY, because it adds clusters aggressively to minimize queueing during spikes
- C
ECONOMY, because it maximizes performance by starting all clusters at the beginning of each hour
- D
STANDARD, because it prevents Snowflake from suspending idle clusters, eliminating all startup delay
Show answer and explanation
Correct answer: A
Explanation
The best choice is STANDARD. In Snowflake multi-cluster warehouses, scaling policies control how aggressively Snowflake starts and stops additional clusters as concurrency changes. STANDARD prioritizes performance and responsiveness, making it the better option when a team wants faster handling of periodic query surges and is willing to accept higher compute cost. ECONOMY is better suited to cost-sensitive workloads where some additional queuing is acceptable. A key exam distinction is that scaling policy applies to multi-cluster warehouses and affects workload-driven cluster behavior; it does not schedule clusters by time of day, nor does it override auto-suspend settings. This aligns with Snowflake documentation on multi-cluster warehouses and scaling policies, where STANDARD is recommended for minimizing queuing and ECONOMY for reducing credit usage.
- A. Correct.
Correct. For multi-cluster warehouses, the STANDARD scaling policy is designed to favor responsiveness by starting additional clusters sooner when workload increases. This generally reduces query queuing and improves performance during bursts of concurrency, though it can increase credit consumption compared to ECONOMY.
- B. Incorrect.
Incorrect. ECONOMY is intended to conserve credits by being more conservative about adding or keeping clusters running. It does not add clusters more aggressively than STANDARD, so it is less appropriate when the goal is minimizing response time during concurrency spikes.
- C. Incorrect.
Incorrect. Scaling policies do not proactively start all clusters based on wall-clock patterns such as the beginning of each hour. Snowflake scales multi-cluster warehouses based on workload and queue conditions, not on a scheduled time trigger implied by the scaling policy.
- D. Incorrect.
Incorrect. STANDARD does not eliminate suspension behavior or guarantee that idle clusters remain running indefinitely. Auto-suspend and cluster lifecycle behavior are separate from the choice between STANDARD and ECONOMY. STANDARD influences how aggressively clusters are added or removed in response to workload, not whether suspension is prevented entirely.