COF-C03 Exam Dumps - SnowPro Core Certification 2026 Exam

Functions like ARRAY_APPEND() and OBJECT_INSERT() are system-defined for manipulating ARRAY and OBJECT types within VARIANT data structures.

Spilling occurs when an operation cannot keep all required intermediate data in warehouse memory. Snowflake first spills data to local storage and can spill additional data to remote storage. Storage access is slower than memory access, so substantial spilling can increase query duration.

Increasing the warehouse size provides more compute resources and more memory per execution cluster. This can reduce or eliminate spilling for a memory-intensive query.

Adding clusters to a multi-cluster warehouse primarily improves concurrency by allowing more queries to execute simultaneously. It generally does not assign multiple clusters to accelerate one individual query.

Warehouse auto-suspend does not terminate an actively running query. Auto-suspend occurs after the warehouse has been inactive for the configured period.

The correct answer is D. Infrastructure management .

Snowflake’s architecture includes the storage layer, compute layer, and Cloud Services layer. The Cloud Services layer coordinates many system-level services required to operate Snowflake.

Why D is correct:

Infrastructure management is handled by the Cloud Services layer. Other Cloud Services responsibilities include authentication, metadata management, query parsing and optimization, access control, and transaction management.

Why the other options are incorrect:

A. Data storage is handled by the database storage layer.

B. File compression is related to storage and data loading/unloading behavior, not the primary Cloud Services responsibility tested here.

C. Query processing is performed by virtual warehouses in the compute layer. Query parsing and optimization happen in Cloud Services, but execution/processing happens in compute.

Official Snowflake documentation reference:

Snowflake documentation describes the Cloud Services layer as managing services such as authentication, infrastructure management, metadata management, query parsing and optimization, access control, and transaction management.

For Parquet files, using the vectorized scanner supports improved schema inference behavior, including use cases involving Iceberg-compatible column definitions.

Corrected typos:

CREATE FIZE FORMAT should be CREATE FILE FORMAT.

VSE_VECTORIZED_SCANNER should be USE_VECTORIZED_SCANNER.

The correct answer is B. GET_PRESIGNED_URL(@ <</b> stage_name > , ' <</b> relative_file_path > ' , [ <</b> expiration_time > ]); .

A pre-signed URL provides temporary access to a staged file. This is useful when users or applications outside Snowflake need access to a file for a limited period of time.

Why B is correct:

GET_PRESIGNED_URL generates a pre-signed URL for a staged file. The URL can be used to access the file without the user needing a Snowflake account, as long as the URL has not expired.

Example:

SELECT GET_PRESIGNED_URL(@my_stage, ' folder/file.csv ' , 3600);

This generates a URL that is valid for a limited time, such as 3600 seconds.

Why the other options are incorrect:

A. BUILD_STAGE_FILE_URL creates a Snowflake file URL, but it does not provide the same temporary external access behavior as a pre-signed URL.

C. GET_RELATIVE_PATH returns the relative path of a staged file, not an access URL.

D. BUILD_SCOPED_FILE_URL creates a scoped URL that is useful within Snowflake-controlled contexts, but for users without Snowflake accounts who need direct temporary access, GET_PRESIGNED_URL is the best answer.

Official Snowflake documentation reference:

Snowflake documentation describes GET_PRESIGNED_URL as a function that returns a pre-signed URL for a staged file, with an optional expiration time.