ADA-C01 Question 395
Single answerApply techniques for cost optimizationA Snowflake administrator is reviewing monthly spend and finds that compute costs increased sharply after several analyst teams began using a shared BI warehouse. Query history shows many short dashboard queries running throughout the day, while warehouse load history shows frequent queueing during peak periods and very low utilization overnight. The teams want better dashboard response times during business hours, but finance has asked the administrator to reduce unnecessary compute spend without changing the BI tool. Which action will best address both requirements?
- A
Increase the warehouse size from MEDIUM to XLARGE and keep a single cluster running 24x7
- B
Convert the warehouse to a multi-cluster warehouse in Auto-scale mode with an appropriate minimum and maximum cluster count, and configure auto-suspend for short idle periods
- C
Disable auto-suspend so the warehouse stays warm and avoids resume latency for dashboard users
- D
Create a dedicated warehouse for each analyst team and leave all warehouses running during business hours
Show answer and explanation
Correct answer: B
Explanation
For cost optimization in Snowflake, administrators should align warehouse configuration with workload characteristics. Concurrency-heavy BI/dashboard workloads with many short-running queries are a classic fit for multi-cluster warehouses in Auto-scale mode. This allows Snowflake to add compute resources only when demand requires it, reducing queueing and improving end-user performance, while avoiding the cost of permanently over-sizing a warehouse. Auto-suspend is another key best practice because warehouses consume credits while running, even when idle. In this scenario, low overnight utilization makes continuous runtime wasteful. Relevant Snowflake documentation and best practices include guidance on selecting warehouse size versus multi-cluster strategy, using Auto-scale for unpredictable concurrency, and enabling auto-suspend/auto-resume to minimize idle compute charges.
- A. Incorrect.
This is not the best cost-optimization approach. Increasing to XLARGE and keeping a single cluster running continuously may reduce some queueing, but it also increases credit consumption significantly and does not address the observed low overnight utilization. A larger always-on warehouse is a common but costly reaction to concurrency problems. Snowflake best practice is to match warehouse behavior to workload patterns rather than simply overprovisioning.
- B. Correct.
This is the best answer. A multi-cluster warehouse in Auto-scale mode is designed for concurrency-heavy workloads such as BI dashboards with many short queries. It can add clusters during peak periods to reduce queueing and improve response time, then scale back down when demand falls. Combining that with auto-suspend for short idle windows helps avoid paying for compute when the warehouse is not in use. This directly addresses both performance during business hours and cost control outside peak demand.
- C. Incorrect.
This is incorrect because disabling auto-suspend usually increases cost, especially when usage is intermittent. While keeping a warehouse warm can reduce resume latency, the scenario specifically highlights low utilization overnight and a need to reduce unnecessary spend. Snowflake warehouses can resume quickly, so leaving them running continuously is generally not the preferred optimization for bursty BI workloads.
- D. Incorrect.
This is a plausible but inefficient approach. Separate warehouses can provide workload isolation, but creating one per team and leaving them all running during business hours often increases total spend and reduces pooling efficiency. The scenario describes a shared BI workload with peak concurrency, which is typically better handled by a properly configured multi-cluster warehouse rather than multiple independently running warehouses.