ADA-C01 Question 412
Select 2Demonstrate cost saving strategiesA Snowflake administrator is asked to reduce monthly compute spend without changing query logic or user-facing SLAs. The environment includes: (1) a dashboard warehouse used heavily from 8 AM to 6 PM and only sporadically overnight, (2) a nightly ELT warehouse that frequently runs large MERGE statements and then sits idle for long periods, and (3) several development warehouses that are often left running after engineers disconnect. Which TWO actions would most directly reduce compute costs while preserving performance for active workloads?
- A
Enable or shorten AUTO_SUSPEND on the dashboard, ELT, and development warehouses, and keep AUTO_RESUME enabled.
- B
Resize the nightly ELT warehouse to a larger size so each MERGE finishes faster, which will always lower total credits consumed.
- C
Configure resource monitors for development warehouses to alert or suspend warehouses after defined credit thresholds are reached.
- D
Disable result cache for the dashboard workload so queries execute on the warehouse instead of returning cached results.
- E
Increase STATEMENT_TIMEOUT_IN_SECONDS for development users so long-running sessions are less likely to terminate.
Show answer and explanation
Correct answers: A, C
Explanation
The best answers are to reduce idle warehouse time and enforce spend controls where usage is less predictable. In Snowflake, warehouse compute charges accrue while the warehouse is running, so AUTO_SUSPEND and AUTO_RESUME are foundational cost-optimization settings, especially for intermittent workloads. Resource monitors provide governance by tracking credit usage and can notify or suspend warehouses when thresholds are exceeded, which is valuable for development or ad hoc usage patterns. By contrast, simply increasing warehouse size is not guaranteed to save money because credit efficiency depends on workload characteristics. Disabling caches would usually increase costs, since Snowflake's persisted query results can avoid unnecessary compute. Relevant Snowflake best practices include warehouse right-sizing, minimizing idle time with AUTO_SUSPEND/AUTO_RESUME, and using resource monitors for cost governance.
- A. Correct.
Correct. Snowflake charges compute credits while a warehouse is running, not only while it is actively executing queries. Using AUTO_SUSPEND with an appropriate short timeout reduces idle warehouse time, and AUTO_RESUME preserves user experience by automatically restarting the warehouse when new queries arrive. This is one of the most common and effective cost-control measures for intermittent workloads such as dashboards outside business hours and development environments.
- B. Incorrect.
Incorrect. Upsizing a warehouse can sometimes improve elapsed runtime, but it does not always reduce total credits consumed. For some workloads, a larger warehouse finishes faster with similar or lower cost, but for others the total cost stays the same or increases. Because the option states this will always lower credits consumed, it is not a reliable cost-saving strategy. It also does not address the stated issue that the ELT warehouse sits idle for long periods after processing completes.
- C. Correct.
Correct. Resource monitors are designed to help control credit usage by tracking consumption and triggering notifications or suspension actions when thresholds are reached. They are particularly useful for development and non-production environments where runaway or forgotten warehouses can drive unnecessary spend. This directly addresses the problem of development warehouses being left running.
- D. Incorrect.
Incorrect. Disabling the result cache would increase compute usage because repeated eligible queries would need to run again on the warehouse instead of returning cached results. Leveraging caching is generally a cost-saving benefit, not something to disable when the goal is to reduce spend without harming user-facing performance.
- E. Incorrect.
Incorrect. Increasing statement timeout affects how long queries are allowed to run before cancellation, but it does not reduce warehouse idle time or directly control warehouse credit consumption. In some cases it could increase spend by allowing inefficient queries to run longer. This does not address the primary cost drivers described in the scenario.