ADA-C01 Question 266
Single answerDifferentiate between user-managed and Snowflake-managed tasks and identify use casesA data engineering team is redesigning nightly and event-driven processing in Snowflake. They currently run a root task every 15 minutes on a dedicated virtual warehouse to check for new files, and several downstream tasks transform and load data only when the root task detects work. The team wants to reduce operational overhead and avoid paying for an always-available warehouse when there is no work to do. However, one downstream task performs a large aggregation that needs a specific warehouse size for predictable performance. Which approach best meets these requirements?
- A
Convert the entire task graph to Snowflake-managed tasks so Snowflake automatically sizes compute for every task, including the large aggregation task.
- B
Keep the root task as a user-managed task on the dedicated warehouse, and convert only the downstream tasks to Snowflake-managed tasks.
- C
Use a Snowflake-managed task for the root task that checks for work, and keep the large aggregation task as a user-managed task on a specified warehouse.
- D
Replace the task graph with a stream on the source table, because streams automatically execute SQL when new data arrives and eliminate the need for task compute.
Show answer and explanation
Correct answer: C
Explanation
Snowflake tasks can run either on user-managed compute, where the task specifies a virtual warehouse, or on Snowflake-managed compute, often referred to as serverless tasks, where Snowflake provisions compute for task execution. Snowflake-managed tasks are a good choice when organizations want to reduce administrative overhead and avoid maintaining a warehouse for intermittent or lightweight scheduled work, such as polling for new data or simple orchestration. User-managed tasks are preferred when administrators need explicit control over the warehouse used, including warehouse size, resource isolation, cost attribution, and predictable performance for heavier transformations. In this scenario, the root task is a good candidate for Snowflake-managed compute because it runs frequently but may often find no work, while the heavy aggregation remains user-managed to ensure a specific warehouse size is used. This aligns with Snowflake best practices for choosing between serverless and warehouse-based task execution.
- A. Incorrect.
Incorrect. Snowflake-managed tasks are useful for serverless execution and reducing warehouse management overhead, but they are not appropriate when a task must run on a specifically chosen warehouse size for predictable performance. A task uses either Snowflake-managed compute or a user-managed warehouse; you would not choose Snowflake-managed for the entire graph if one task has a clear requirement for dedicated, explicitly sized compute.
- B. Incorrect.
Incorrect. This keeps the polling/checking task on a dedicated warehouse, which does not satisfy the goal of avoiding warehouse cost and administration when there is no work to do. In addition, child tasks in a task graph do not need their own warehouse if they are Snowflake-managed, but retaining the root polling task as user-managed preserves the main inefficiency the team is trying to remove.
- C. Correct.
Correct. A Snowflake-managed task is a strong fit for the root task that periodically checks whether work exists, because it removes the need to provision and manage a warehouse for lightweight orchestration logic. The large aggregation task is better as a user-managed task because the team wants a specific warehouse size and predictable performance characteristics. This design appropriately differentiates serverless task use cases from cases where explicit warehouse control is preferred.
- D. Incorrect.
Incorrect. Streams track change data capture metadata and are commonly paired with tasks, but they do not execute SQL or replace task scheduling/orchestration by themselves. This option reflects a common misconception that streams are event processors. In practice, a task still provides the execution mechanism, whether user-managed or Snowflake-managed.