COF-C03 Question 273
Single answerWorkload management best practicesA company uses a single MEDIUM virtual warehouse for hourly ELT jobs, interactive BI dashboards, and ad hoc analyst queries. At the top of each hour, dashboard users report slow performance because the ELT jobs start at the same time and consume most of the warehouse resources. The company wants to improve dashboard responsiveness during these peak periods while minimizing cost increases and avoiding major application changes. Which solution BEST follows Snowflake workload management best practices?
- A
Enable multi-cluster on the existing warehouse in Maximized mode so Snowflake can add clusters when concurrency increases
- B
Create a separate warehouse for BI workloads and keep the ELT jobs on the current warehouse
- C
Resize the existing warehouse from MEDIUM to 2XLARGE so all workloads have more compute available
- D
Enable Query Acceleration Service on the existing warehouse so queued dashboard queries bypass ELT resource contention
Show answer and explanation
Correct answer: B
Explanation
The best answer is to create a separate warehouse for BI workloads. In Snowflake, virtual warehouses provide compute isolation, which is one of the main workload management advantages of the platform. When interactive BI and batch ELT share the same warehouse, user-facing queries can slow down during heavy ETL/ELT processing. Snowflake best practices recommend separating workloads with different performance and usage patterns into different warehouses so they can scale independently, be suspended independently, and be tuned for their own SLAs and cost profiles. Multi-cluster warehouses are useful when many concurrent queries on the same workload type cause queueing, but they are not the primary solution for isolating batch and interactive workloads. Similarly, simply resizing a warehouse can increase cost without fully resolving resource competition. See Snowflake documentation on virtual warehouses, multi-cluster warehouses, and workload isolation best practices.
- A. Incorrect.
Incorrect. Multi-cluster warehouses are primarily intended to handle higher concurrency by adding clusters when many queries are running at the same time. In this scenario, the main issue is competing workload types: resource-intensive ELT jobs are contending with user-facing BI queries. Multi-cluster can help with queueing caused by concurrency, but it does not provide the same isolation as separating workloads into different warehouses. Snowflake best practice is to isolate different workload types when predictable performance is required.
- B. Correct.
Correct. Separating BI and ELT onto different virtual warehouses is a core Snowflake workload management best practice. Virtual warehouses are independent compute clusters, so placing dashboard queries on their own warehouse isolates them from batch ELT activity and improves performance consistency. This approach also requires minimal application change because queries can be routed to a different warehouse, and it avoids overprovisioning one shared warehouse for all workloads.
- C. Incorrect.
Incorrect. Increasing the warehouse size may improve performance temporarily, but it does not address the root cause of mixed workloads competing for the same compute resources. It can also significantly increase cost. Snowflake best practices generally favor workload isolation using separate warehouses before simply scaling up a shared warehouse that supports very different usage patterns.
- D. Incorrect.
Incorrect. Query Acceleration Service can improve the performance of eligible individual queries, especially parts of queries with heavy scan and filter processing, but it is not designed to solve general workload isolation or resource contention between ELT and BI users on a shared warehouse. It also does not make queued queries bypass contention in the way this option suggests.