COF-C03 Question 4
Single answerCloud Services layerA data engineering team notices that a Snowflake account is spending significantly more on Cloud Services even during periods when no large ETL jobs are running. Investigation shows that several BI tools and custom applications repeatedly submit short metadata-heavy queries such as SHOW commands, DESCRIBE statements, and frequent queries against INFORMATION_SCHEMA. The team wants to reduce unnecessary Cloud Services consumption without affecting core data processing workloads. Which action is the MOST effective?
- A
Increase the size of the virtual warehouses used by the BI tools so metadata queries complete faster
- B
Reduce the frequency of metadata polling and cache object metadata in the applications where appropriate
- C
Convert all permanent tables to transient tables so fewer metadata operations are required
- D
Move the workload to a multi-cluster warehouse so Cloud Services load is distributed automatically
Show answer and explanation
Correct answer: B
Explanation
The Cloud Services layer in Snowflake manages functions such as authentication, access control, metadata management, query parsing, and query optimization/coordination. In real environments, high volumes of short control-plane requests from dashboards, orchestration tools, and custom apps can increase Cloud Services consumption even when warehouse-intensive ETL is minimal. The best mitigation is to reduce unnecessary metadata requests at the source, for example, by lowering polling frequency, avoiding excessive INFORMATION_SCHEMA scans, and caching metadata when the application design allows it. Snowflake documentation describes the Cloud Services layer as the set of services that coordinate activities across the platform, separate from the virtual warehouse compute layer used for data processing. This is why simply resizing or scaling warehouses does not effectively solve metadata-driven Cloud Services cost patterns.
- A. Incorrect.
Incorrect. Cloud Services activities such as authentication, metadata management, access control checks, and query parsing/optimization are not primarily reduced by making a warehouse larger. Many metadata-oriented requests, including SHOW and DESCRIBE operations, rely heavily on the Cloud Services layer rather than warehouse compute. Increasing warehouse size may increase compute cost without addressing the root cause.
- B. Correct.
Correct. The scenario points to excessive metadata and control-plane activity: repeated SHOW, DESCRIBE, and INFORMATION_SCHEMA access from BI tools and applications. The Cloud Services layer is responsible for services such as authentication, infrastructure management, metadata management, and query optimization/coordination. Reducing unnecessary metadata polling and caching metadata where feasible directly lowers repeated Cloud Services requests while preserving warehouse resources for actual data processing.
- C. Incorrect.
Incorrect. Table type selection such as permanent versus transient affects data protection and Fail-safe behavior, not whether applications repeatedly issue metadata requests. Converting table types would not meaningfully reduce Cloud Services usage caused by frequent catalog and schema introspection.
- D. Incorrect.
Incorrect. Multi-cluster warehouses address concurrency for compute execution on warehouse workloads. They do not offload or distribute Cloud Services layer consumption in the way described here. Metadata-heavy commands are not solved by adding warehouse clusters, so this option reflects a common misconception that all Snowflake workload costs scale through warehouses.