ARA-C01 Question 344
Select 3Impact of costsA global retail company runs a Snowflake-based analytics platform for finance, merchandising, and data science teams. The architect notices that monthly spend has increased significantly even though the amount of data loaded each day has remained stable. Investigation shows that several dashboards run every 5 minutes, ad hoc analyst queries frequently scan large fact tables, and a data science team occasionally uses large warehouses for model feature generation. The company wants to reduce cost without materially affecting business SLAs. Which TWO actions would most effectively reduce Snowflake compute cost in this scenario?
- A
Enable and tune auto-suspend/auto-resume on warehouses, and isolate workloads so infrequent data science jobs do not keep BI warehouses running
- B
Create additional virtual warehouses for different workloads, because more warehouses inherently reduce total Snowflake cost
- C
Use Query Profile and warehouse-level monitoring to identify inefficient queries and right-size warehouses for dashboard, ad hoc, and data science workloads
- D
Increase the Time Travel retention period on all tables so users can rerun historical queries more efficiently
- E
Use materialized views or other precomputation strategies selectively for repeatedly executed, expensive dashboard queries after validating that maintenance cost is justified
Show answer and explanation
Correct answers: A, C, E
Explanation
The best answers are options 1, 3, and 5 because the scenario is specifically about reducing Snowflake compute cost while maintaining SLAs. In Snowflake, cost impact is commonly driven by warehouse runtime, warehouse size, concurrency strategy, and query efficiency. Auto-suspend/auto-resume reduces idle compute charges, and workload isolation prevents one team's usage pattern from causing another workload to require a larger or longer-running warehouse. Query tuning and right-sizing are core best practices because poorly optimized SQL and oversized warehouses often drive avoidable spend. For highly repetitive dashboard workloads, selective precomputation such as materialized views can reduce repeated execution costs, but architects must evaluate maintenance overhead carefully. Option 2 reflects a common misconception: more warehouses are not automatically cheaper. Option 4 addresses storage-related behavior and can increase storage charges without solving the compute issue described. These recommendations align with Snowflake best practices around warehouse management, query profiling, workload isolation, and evaluating the trade-offs of performance optimization features.
- A. Correct.
Correct. Compute charges in Snowflake are driven primarily by virtual warehouse usage. Auto-suspend and auto-resume help avoid paying for idle compute time, especially for intermittent workloads. Isolating workloads into separate warehouses can also prevent a sporadic large job from keeping a shared BI warehouse active or causing contention that leads teams to scale up unnecessarily. This is a common architectural cost-control practice.
- B. Incorrect.
Incorrect. Creating more warehouses does not inherently reduce cost. Separate warehouses can improve workload isolation and sometimes reduce waste, but they can also increase cost if each warehouse remains active, is oversized, or is left running unnecessarily. The benefit comes from targeted isolation and proper sizing, not from simply increasing the number of warehouses.
- C. Correct.
Correct. Query optimization and right-sizing are among the most effective ways to reduce compute spend. If dashboards are executed every 5 minutes and ad hoc users are scanning large tables, architects should use Query Profile, query history, and warehouse monitoring to identify excessive scans, poor pruning, unnecessary joins, and warehouses that are too large for the workload. Smaller or better-matched warehouses, along with more efficient SQL, can materially reduce credits consumed.
- D. Incorrect.
Incorrect. Increasing Time Travel retention generally increases storage cost, not reduces compute cost. It does not make standard historical query execution inherently more efficient for this scenario. This option confuses storage-related features with compute cost optimization.
- E. Correct.
Correct. For repetitive, expensive queries such as dashboard refreshes, selective precomputation can reduce repeated compute consumption at query time. Materialized views can improve performance and reduce repeated work for suitable patterns, but they have maintenance costs because Snowflake must keep them updated. Therefore, they should be used only where the net cost/performance trade-off is favorable.