ADA-C01 Question 382
Single answer4.5 Manage and optimize costs.A Snowflake administrator is reviewing monthly costs and notices that compute charges increased sharply after several analytics teams began using the platform. Query history shows many short, intermittent dashboard queries running every few minutes throughout the day, and warehouse monitoring shows frequent warehouse resumes. The teams also use a separate warehouse for a nightly 3-hour ETL job. The administrator needs to reduce compute cost without significantly affecting dashboard responsiveness or ETL completion time. Which action should the administrator take?
- A
Move both the dashboard workloads and the nightly ETL job onto a single larger warehouse so all compute is centralized and easier to manage
- B
Increase the auto-suspend time on the dashboard warehouse so it stays running longer between intermittent queries, while leaving the ETL warehouse configuration unchanged
- C
Convert the dashboard warehouse to a multi-cluster warehouse with maximum clusters set to 3 to reduce queueing and lower credits consumed
- D
Disable auto-suspend on the dashboard warehouse to avoid repeated resume events and eliminate extra compute charges
Show answer and explanation
Correct answer: B
Explanation
The key cost optimization principle in this scenario is matching warehouse behavior to workload patterns. Snowflake warehouses incur a minimum billing charge when started, so workloads with short, intermittent query bursts can become inefficient if warehouses suspend too aggressively and then resume repeatedly. For dashboard or BI workloads that run every few minutes, a somewhat longer auto-suspend setting often reduces unnecessary resume cycles and improves user experience. By contrast, the nightly ETL warehouse already runs for a sustained period, so its configuration may not need adjustment.
Snowflake best practices for cost management include monitoring warehouse utilization, separating workloads for performance isolation, and tuning warehouse size and auto-suspend/auto-resume settings based on query patterns. Multi-cluster warehouses are useful for managing high concurrency, but they are not a primary tool for reducing resume-driven costs. Likewise, disabling auto-suspend is generally not recommended unless there is a clear operational reason, because idle running time consumes credits unnecessarily.
Relevant Snowflake documentation and best practices include guidance on virtual warehouse billing, auto-suspend and auto-resume behavior, and using warehouse monitoring/QUERY_HISTORY to identify inefficient compute patterns.
- A. Incorrect.
Incorrect. Consolidating intermittent dashboard queries and a long-running ETL workload onto one warehouse can create resource contention and does not inherently reduce cost. In many real environments, separating workloads is a best practice for both performance isolation and cost control. A larger warehouse may also consume more credits per hour than necessary for either workload individually.
- B. Correct.
Correct. Snowflake bills warehouses with a minimum charge when they resume, so very short, frequent query bursts can lead to inefficient credit usage if the warehouse repeatedly suspends and resumes. For intermittent dashboard traffic, increasing the auto-suspend threshold can reduce resume frequency and improve credit efficiency while preserving responsiveness. Keeping the ETL warehouse unchanged is reasonable because the ETL job already runs continuously for several hours and is less likely to suffer from resume-related inefficiency.
- C. Incorrect.
Incorrect. Multi-cluster warehouses are primarily used to handle concurrency by automatically adding clusters when query demand increases. They do not reduce cost for a workload characterized mainly by short, intermittent activity with frequent resumes. In fact, enabling multiple clusters can increase compute consumption if extra clusters are started. The issue described is warehouse start/stop behavior, not queueing pressure.
- D. Incorrect.
Incorrect. Disabling auto-suspend entirely would likely increase costs because the dashboard warehouse would continue running during idle periods throughout the day and possibly overnight. Although this avoids resume events, it usually wastes more credits than tuning auto-suspend appropriately. The misconception is assuming resume events themselves create a larger cost penalty than prolonged idle runtime.