SnowPro Advanced Architect flashcards

32 cards. Read the term and the definition here. The deck at the top of this page is optional.

Cloud services layer

Snowflake layer handling metadata, query parsing and optimization, access control, and transactions. It bills credits only when daily usage exceeds 10% of warehouse credits.

Micro-partition

Immutable columnar storage unit of 50 to 500 MB uncompressed. Per-column minimum and maximum values stored in metadata enable partition pruning during queries.

Multi-cluster warehouse

Enterprise feature that starts and stops additional compute clusters automatically to absorb concurrent queries. Each running cluster is billed credits separately for its runtime.

Fail-safe

Fixed seven-day period beginning after Time Travel retention ends. Data is recoverable only by Snowflake support and is not queryable by users or roles.

Managed access schema

Schema in which the schema owner, not individual object owners, grants privileges on contained objects. Future grants are managed centrally at schema level.

Network policy

Allow list or block list of IP addresses enforced at account or user level. A user-level policy overrides the account policy for that specific user.

Tag-based masking

Enterprise feature attaching a masking policy to a tag. Any column assigned that tag is masked automatically without editing policy assignments column by column.

Copy current grants

Clause on GRANT OWNERSHIP that transfers an object to a new owner role while preserving existing privilege grants instead of revoking them.

Result cache

Account-level cache of query results reused for 24 hours when identical query text runs and underlying micro-partitions are unchanged, avoiding warehouse execution.

Clustering depth

Average number of overlapping micro-partitions that a value's rows occupy. A depth near one indicates good clustering and effective pruning.

Search optimization service

Enterprise feature building an auxiliary index to accelerate equality, IN, and substring predicates and VARIANT field access. It is billed as serverless compute.

Remote spill

Bytes written to remote storage when a query exceeds warehouse memory and local disk. It signals an undersized warehouse and usually inflates runtime substantially.

Zero-copy clone

Metadata-only copy of a database, schema, or table. Clone and source share micro-partitions until either side modifies data, adding storage only for changes.

Row access policy

Enterprise feature mapping roles to a boolean expression that filters rows at query time. A mapping table joined inside the policy handles many groups.

ACCESS_HISTORY view

ACCOUNT_USAGE view recording which columns each query read or wrote. It supports column-level lineage and auditing of sensitive data access.

UNDROP TABLE

Command restoring a dropped table from Time Travel without cloning. The table returns with its original name and, where possible, original grants.

Secure data sharing

Provider grants a share and consumer creates a database from it. No data is copied and consumer queries run on consumer warehouse credits.

Reader account

Account created and funded by a provider so a consumer without a Snowflake account can query shared data. The provider bears both storage and compute cost.

Secure view

View whose definition and underlying objects are hidden from consumers. It is the standard way to expose filtered or column-limited data in a share.

Database replication

Copies a primary database to secondary accounts in other regions or clouds. It requires Business Critical or higher and underpins failover and read access.

Stream

Object recording change-tracking offsets on a table, view, or external table so downstream tasks consume inserted, updated, and deleted rows exactly once.

Task DAG

Set of tasks linked by dependencies with a single root task. Child tasks run after predecessors succeed and can execute on different warehouses.

Dynamic table

Object defined by a query plus a target lag, refreshed automatically using incremental or full refresh. It replaces many custom streams-and-tasks pipelines.

Snowpipe Streaming

API and Kafka connector that inserts rows directly into tables with seconds-level latency, skipping file staging. Its compute is serverless and billed by usage.

Storage integration

Account-level object storing cloud provider credentials and permitted locations. External stages reference it instead of embedding access keys directly.

Validation mode

COPY INTO option that loads data and then rolls back, reporting format and transformation errors without committing any rows to the target table.

Snowpipe auto-ingest

Continuous loading triggered by cloud event notifications feeding a queue that Snowflake reads. Serverless compute loads each new file shortly after arrival.

External table

Read-only table over files in an external stage. Data is not loaded, and partitions depend on metadata synced by events or manual refresh.

Resource monitor

Object with a credit quota and notify or suspend actions, set at account or warehouse level, evaluated over a daily, weekly, monthly, or never-resetting interval.

Warehouse credit rate

Credits consumed per hour scale with size: X-Small 1, Small 2, Medium 4, Large 8, X-Large 16, doubling with each further size step.

Cloud services billing

Cloud services credits are uncharged until they exceed 10% of that day's warehouse credits. Only the amount above that threshold appears on the bill.

Metering daily history

ACCOUNT_USAGE view aggregating credits by service type and day. It attributes consumption across warehouses, serverless features, and storage-related services.