Snowflake SnowPro Core (COF-C02) Practice Test

COF-C02 practice test: what's covered

• Snowflake
• Architecture
• Virtual Warehouses
• Micro-partitions
• Loading
• Sharing

Free COF-C02 sample questions

1. A data team complains that a heavy reporting workload slows down the nightly ETL load that runs at the same time. Both run in the same database. Which property of Snowflake's architecture lets you fix this WITHOUT copying or moving the data?

   • A. Snowflake stores each table on a dedicated server, so you must shard the table across nodes
   • B. Storage and compute are separated, so you can point a SEPARATE virtual warehouse at the same data — each workload gets its own isolated compute
   • C. You must enable a read replica of the database in another region
   • D. You must create a clustering key so the two jobs read different micro-partitions
   Show answer

   Correct: B. Snowflake's multi-cluster shared-data architecture decouples storage from compute. Many independent virtual warehouses can query the SAME data concurrently with zero contention — give ETL its own warehouse and reporting another. The trap (B-style misconception) of dedicated per-table servers/sharding is the old MPP model Snowflake replaced. Read replicas (C) and clustering keys (D) do not isolate compute.

2. Which layer of Snowflake's three-layer architecture is responsible for query parsing, optimization, transaction management, metadata, security, and access control?

   • A. The Database Storage layer
   • B. The Cloud Services layer
   • C. The Query Processing (compute) layer
   • D. The Storage Integration layer
   Show answer

   Correct: B. Snowflake has three layers: Database Storage (compressed columnar data in cloud blob storage), Query Processing (virtual warehouses = compute), and Cloud Services (the "brain" — optimizer, metadata, transactions, authentication, RBAC, infrastructure management). Query optimization and metadata live in Cloud Services, not in the warehouses. "Storage Integration" is an object for external stages, not an architecture layer.

3. Which THREE statements correctly describe how Snowflake physically organizes table data in micro-partitions? (Choose 3)

   • A. Data is automatically divided into immutable micro-partitions of roughly 50–500 MB of uncompressed data each
   • B. Each micro-partition is stored in compressed columnar format
   • C. Snowflake captures metadata (e.g. min/max values) per micro-partition to enable pruning
   • D. The administrator must manually create and manage fixed-size partitions
   • E. Users create B-tree indexes on micro-partitions to speed lookups
   • F. Micro-partitions are mutable and rewritten in place on every UPDATE
   Show answer

   Correct: A and B and C. Snowflake automatically and transparently stores table data in immutable micro-partitions (~50–500 MB uncompressed), written in compressed columnar format, with per-partition metadata (min/max, counts) that powers pruning. Users never manage partitions manually and there are no user-created B-tree indexes (the common misconceptions). Micro-partitions are immutable — changes write new partitions rather than rewriting in place.

4. Which THREE types of cache exist in Snowflake and improve performance? (Choose 3)

   • A. Metadata cache (in the Cloud Services layer)
   • B. Result cache (query results reused for 24 hours when the query and data are unchanged)
   • C. Warehouse cache / local disk cache (data cached on the virtual warehouse's SSD)
   • D. Browser cache stored on the end-user's laptop
   • E. A per-table index cache that the user must build with CREATE INDEX
   • F. A replication cache that mirrors data to every region
   Show answer

   Correct: A and B and C. Snowflake has three caches: (1) metadata cache in Cloud Services, (2) result cache that returns identical results for up to 24 hours without using a warehouse, and (3) the warehouse/local disk cache on the running warehouse's SSD. There is no user-built index cache (Snowflake has no traditional indexes) and no "replication cache."

5. A query is re-run with identical SQL text, and the underlying table has not changed since the previous run 10 minutes ago. The warehouse is currently suspended. What happens?

   • A. The warehouse must resume and fully re-scan the data
   • B. The result is served from the result cache WITHOUT resuming a warehouse, consuming no compute credits
   • C. The query fails because the warehouse is suspended
   • D. Snowflake automatically provisions a new warehouse to recompute the result
   Show answer

   Correct: B. The result cache is held in the Cloud Services layer and can satisfy an identical query (same SQL, same role privileges, unchanged data) for up to 24 hours without resuming any warehouse — so no credits are consumed. The misconception is that a suspended warehouse must always resume; for a result-cache hit it does not.

Last updated: 2026-10-19

COF-C02 practice test FAQ