ARA-C01 Question 393
Single answerScale in/out (multi-cluster warehouse/auto-scaling)A retail company runs thousands of short BI queries during business hours. The queries are submitted by many concurrent users through dashboards, and users frequently complain that response times spike even though individual queries are not compute-intensive. The current warehouse is a single-cluster MEDIUM warehouse with AUTO_SUSPEND enabled. An architect must improve concurrency during peak hours while minimizing unnecessary cost during quieter periods. Which configuration change is the best fit for this requirement?
- A
Resize the warehouse from MEDIUM to 2X-LARGE and keep it as a single-cluster warehouse
- B
Convert the warehouse to a multi-cluster warehouse in AUTO-SCALE mode with an appropriate minimum and maximum cluster count
- C
Disable AUTO_SUSPEND so the existing single-cluster warehouse remains warm and avoids startup latency
- D
Enable Query Acceleration Service on the warehouse to eliminate queueing from concurrent dashboard traffic
Show answer and explanation
Correct answer: B
Explanation
The best answer is to use a multi-cluster warehouse in AUTO-SCALE mode. In Snowflake, scaling up increases the size of a single cluster and is most useful when individual queries need more compute. Scaling out with a multi-cluster warehouse addresses concurrency by allowing multiple clusters to process separate queries in parallel, which is particularly effective for many short-running BI/dashboard queries. AUTO-SCALE helps balance performance and cost by adding clusters only when needed and removing them as demand subsides. This aligns with Snowflake best practices for high-concurrency workloads. Relevant Snowflake documentation describes multi-cluster warehouses as the preferred approach for supporting concurrent users and reducing queueing, especially for dashboard and reporting workloads, while warehouse resizing is more appropriate for improving the performance of individual heavy queries.
- A. Incorrect.
Incorrect. Scaling up a warehouse increases resources available to each running query, which can help with large or compute-heavy workloads. However, this scenario is primarily about high concurrency from many short BI queries. A larger single cluster does not address concurrency as effectively as scaling out with multiple clusters, so users may still experience queueing during peaks while costs increase significantly.
- B. Correct.
Correct. Multi-cluster warehouses are designed to improve concurrency by adding additional clusters when demand increases. In AUTO-SCALE mode, Snowflake can start additional clusters as query load rises and reduce them when demand falls, making this the best option for handling bursty dashboard traffic while controlling cost. Setting appropriate minimum and maximum cluster counts aligns with the requirement to improve peak-hour performance without overpaying during quieter periods.
- C. Incorrect.
Incorrect. Disabling AUTO_SUSPEND may reduce cold-start delays, but it does not solve the main issue: queueing caused by many concurrent requests on a single cluster. This change would likely increase cost because the warehouse would continue running when idle, without materially improving concurrency under peak load.
- D. Incorrect.
Incorrect. Query Acceleration Service is intended to improve the performance of eligible individual queries by offloading portions of scan and filter processing, not to solve warehouse-level concurrency bottlenecks caused by many simultaneous dashboard requests. It is not the primary mechanism for eliminating queueing in this scenario.