Free ARA-C01 practice test: 30 questions with answers

An External OAuth security integration lets Snowflake accept access tokens issued by the customer's identity provider. Network policies restrict source addresses, masking policies hide values, and resource monitors track credits.

CURRENT_ROLE returns the active role of the session, which a CASE expression in the masking policy can compare with HR_ANALYST. The other functions return user, account or task information, not the role that decides visibility.

A tag-based masking policy applies to every object carrying the tag, including new columns, which keeps protection consistent. Manual changes and row policies do not scale this way, and exporting data moves it outside the governed platform.

TLS protects data in transit between clients and Snowflake, and encryption at rest protects stored data. Shared passwords, unencrypted stages and disabled certificate validation each remove a protection rather than adding one.

PrivateLink or Azure Private Link routes traffic over the cloud provider's private network to Snowflake, and network policies add source controls. Public endpoints with allow lists or MFA still traverse the public internet, and laptop VPNs do not cover system-to-system traffic.

ACCESS_HISTORY records which objects and columns each query accessed, which supports column-level audits. The other sources describe table metadata, privileges or storage, not which queries read a column.

The WHEN clause with SYSTEM$STREAM_HAS_DATA skips runs when the stream is empty, which avoids idle warehouse use. A schedule alone runs regardless of data, and AFTER links tasks rather than streams, while WAIT FOR is not a task clause.

Snowpipe Streaming sends rows straight to the table, while file-based Snowpipe reacts to files staged in cloud storage. Snowpipe is serverless and does not need a dedicated warehouse per file, it is not limited to JSON, and file-based Snowpipe commonly uses storage event notifications.

COPY INTO tracks loaded files in load metadata and skips them unless FORCE is set to TRUE, which makes retries idempotent by default. FORCE = TRUE does the opposite, PURGE deletes source files, and CONTINUE only affects error rows.

MERGE matches on the business key, so the correction updates the existing row or inserts a new one. Insert-only designs push duplicate handling into every consumer, and rebuilds are expensive and disruptive.

Partitioning on folder-derived columns lets queries read only the relevant files, and refreshed metadata keeps the table aware of new files. Compressed CSV and single large files reduce parallelism, and stopping refresh leaves new data invisible.

AFTER defines a predecessor relationship in a task graph, so the child runs when the root finishes. The other clauses are not valid task dependency syntax in Snowflake.

VARIANT keeps the raw semi-structured payload intact, and promoting the hot attributes to typed columns preserves performance for common queries. BLOB and large VARCHAR hide structure, and one table per attribute makes ingestion and queries fragile.

Clustering on the frequently filtered date column co-locates similar values so pruning eliminates more micro-partitions. A unique surrogate key gives no pruning benefit for range filters, a column no query uses is wasted cost, and automatic layout does not match the workload.

A Type 2 pattern keeps each version with effective and end dates, and a MERGE driven by a stream applies the changes incrementally. Overwriting or reloading without dates destroys history, and external CSV history is not queryable as dimension versions.

Zero-copy clones share storage until data diverges, and secure views expose one base table to several audiences without copies. Full copies, per-region duplicates and manual CSV exports multiply storage and create drift.

A row access policy enforces tenant filtering on every query against the shared table, and secure views give each consumer a controlled interface. Warehouses control compute rather than data visibility, and files in a stage do not restrict query results.

Hybrid tables are built for low-latency transactional row operations and can be analyzed alongside standard tables. Transient and temporary tables are for storage cost or session scope, and external tables are read-only views of files in cloud storage.

Right-sizing with auto-suspend stops idle billing, and separate warehouses isolate workloads so each is sized for its own needs. Oversized always-on warehouses and disabling suspend increase cost, and a single saturated warehouse slows critical jobs.

The search optimization service builds access paths for selective lookups on high-cardinality columns. Clustering on a random column adds maintenance cost without pruning benefit, cache tuning does not help first-time queries, and a larger warehouse does not fix the access path.

A dynamic table can contain joins and maintains results to a target lag, which fits this requirement. A standard view recomputes on each query, materialized views have restrictions on joins, and external tables read files in place.

Query acceleration offloads parts of a large scan to additional serverless compute, which suits this profile. More clusters help concurrency, disabling the cache removes reuse, and clustering on an unused column provides no benefit.

A multi-cluster warehouse scales between minimum and maximum cluster counts as concurrency rises and falls. The other options are not valid warehouse settings for this, and clustering keys belong to tables.

Filters on clustered or partitioning columns let Snowflake prune micro-partitions, and search optimization speeds selective lookups. SELECT * reads more columns, wrapping filter columns in functions can prevent pruning, and statistics help the planner.

Failover groups replicate selected databases, roles and other objects to a secondary account or region and let the secondary be promoted during an outage. Clones are point-in-time copies in the same account, shares do not provide recovery, and Time Travel covers history rather than regional failover.

A reader account is a Snowflake account the provider creates and manages so that a consumer without its own account can use shared data. Exports create copies, shared admin credentials are unacceptable, and public stage links expose files without access control.

Replication creates a separate copy for recovery or locality, and a share exposes provider data read-only without duplicating it. Shares do not copy data into the consumer, replication works across regions, and consumers cannot change provider tables through a share.

A secure view hides its definition and limits the optimizer's exposure of underlying data, which suits shared data. A standard view exposes its definition to users with access, and materialized views and dynamic tables are storage and refresh objects.

Failback requires replicating the changes made during the outage back to the original region, verifying it is current, and then promoting it. Recreating from scratch or restoring old exports loses changes, and running two unsynchronized primaries causes divergence.

The recovery point objective is the maximum tolerable age of recovered data, which is the amount of data that may be lost. Recovery time objective measures how long restoration may take, and the other terms do not describe data loss.