Snowflake Exam Sample Questions: Core, Associate & Advanced Explained

正解: A, C

The Cloud Services Layer is Snowflake's "brain" — it handles query parsing, optimization, and metadata management (A), plus authentication, RBAC, and access control (C). Option B (physical data storage) belongs to the Storage Layer (micro-partitions on cloud object storage). Option D (query compute) belongs to the Compute Layer (Virtual Warehouses). Cloud Services usage is normally free, but you are billed only if daily Cloud Services usage exceeds 10% of total compute usage.

正解: B

In Auto-scaling mode, a multi-cluster Warehouse automatically scales out when queries queue up and scales back in when query load drops (B is correct). This solves the concurrency problem. Option A is wrong: multi-cluster Warehouses require Enterprise Edition or above. Option C is also wrong: more clusters increase concurrency capacity, but to speed up an individual query you scale up the Warehouse size. In Maximized mode (option D), the configured number of clusters start, but the Warehouse's auto-suspend setting still applies — so they don't necessarily run nonstop.

正解: B

Time Travel uses the AT or BEFORE clause to specify a past point in time. For a timestamp-based query you use AT(TIMESTAMP => <timestamp>), and you compute the 30-days-ago timestamp with DATEADD(day, -30, CURRENT_TIMESTAMP()) — so B is correct. Option A's OFFSET takes seconds, but the expression is malformed (you would pass -30*86400 as a plain number). Option C's "DAYS => 30" is not valid Time Travel syntax. Option D's "TRAVEL TO" simply doesn't exist in Snowflake. To access data from 30 days ago you also need Enterprise Edition or above with DATA_RETENTION_TIME_IN_DAYS set to 30 or more.

正解: C

ACCOUNTADMIN sits at the top of Snowflake's system-defined role hierarchy and inherits the privileges of both SYSADMIN and SECURITYADMIN. It can manage databases, Warehouses, users, roles, and all security settings. SYSADMIN is focused on creating and managing databases and Warehouses, SECURITYADMIN on managing roles, privileges, and user security, and USERADMIN on creating users and roles. Snowflake best practice is to minimize day-to-day use of ACCOUNTADMIN and rely on SYSADMIN or custom roles instead.

正解: B

When sharing data with an external partner who has no Snowflake account, a Reader Account is the right design. A Reader Account is created and managed by the data provider, and the partner can run SQL queries against the shared data with no Snowflake contract of their own. Compute costs for the Reader Account are borne by the provider. File export (A) is inferior to Secure Data Sharing on freshness and security. Forcing the partner onto an Enterprise contract (C) imposes unnecessary cost. A public listing (D) is inappropriate for a targeted share to a specific partner.

正解: A

To expand a VARIANT array or object into rows in Snowflake, you combine the FLATTEN function with the TABLE function. TABLE(FLATTEN(input => data)) expands the array into rows, and each expanded element is accessed via f.value. Accessing JSON keys with colon notation (f.value:type) is Snowflake-specific syntax. UNNEST is a PostgreSQL/BigQuery function (B), EXPLODE is a Spark SQL function (C), and JSON_EXTRACT is a MySQL/BigQuery function (D) — none of them work in Snowflake. In Snowflake you access semi-structured data with colon notation or the GET_PATH function.