SnowPro Advanced Architect key terms

Immutable, compressed columnar storage unit holding roughly 50 to 500 MB of uncompressed table data that Snowflake creates automatically; each stores metadata such as column min/max values used for pruning.

Snowflake layer that handles authentication, access control, metadata, query parsing and optimization, and transaction coordination; it stores no table data and does not execute user queries.

Independently sized cluster of compute nodes (XSMALL through 6XL) that runs queries and DML. Storage is separate, so a warehouse can be resized, suspended, or dropped without affecting data.

Warehouse configuration that adds clusters automatically when queries queue and removes them when demand falls. Modes are Maximize and Auto-scale, and each running cluster bills separately.

Graphical breakdown of a completed query's execution plan showing each operator's rows, time, bytes, and spill, used to find bottlenecks such as exploding joins, poor pruning, or excessive spilling.

Optimization in which Snowflake reads micro-partition metadata to skip partitions that cannot hold rows matching a filter, cutting scanned bytes for selective predicates. Clustering keys make it more effective.

Optional column set or expression on a table that Snowflake uses to co-locate similar rows in the same micro-partitions, improving pruning for very large tables filtered on those columns.

Service that maintains an auxiliary search access path so selective equality lookups, substring searches, and predicates such as CONTAINS or STARTSWITH avoid full scans; enabled per table or schema.

Service that offloads scan-heavy portions of eligible queries to shared compute, reducing the impact of outlier queries. It is enabled on a warehouse with a scale factor that limits added cost.

Query results held in the Cloud Services layer and reused when identical query text is submitted and underlying data is unchanged; valid 24 hours and extended while reused, consuming no warehouse credits.

SSD storage on warehouse nodes holding table data read from remote storage. It survives while the warehouse runs, is lost on suspend, and speeds repeat queries without extra cost.

Retention of historical table, schema, and database states for up to 90 days depending on edition and setting, allowing AT or BEFORE queries and restoration of objects after mistakes.

Non-queryable seven-day period following Time Travel during which only Snowflake Support can recover data after a catastrophic loss; users cannot query it and it cannot be disabled or shortened.

Creates a table, schema, or database that initially shares micro-partitions with the source instead of copying bytes. Later changes to either object are tracked separately, so storage grows only for changes.

Masking policy attached to a column that rewrites values at query time depending on the role in the session, so unauthorized users see masked results while stored data stays intact.

Schema-level policy returning a boolean per row, evaluated at query time to hide rows the current role should not see. It applies to each table or view it is attached to.

Assigning key-value tags to Snowflake objects for classification, cost attribution, and governance. Tags are queryable in account usage views and can drive tag-based masking policies.

A view whose definition and underlying objects are hidden from users who query it, so consumers can use shared data without seeing the SQL logic or the base tables.

Directed acyclic set of roles in which each role inherits privileges of every role granted to it. A session's active role determines which objects the user can access.

Role representing a business function or job family that is granted to users, usually sitting above access roles that hold the actual object privileges.

Account- or user-level rule that allows or blocks connections by IP address range, and can require MFA or restrict logins to private connectivity endpoints.

Passwordless authentication using an RSA key pair: the public key is stored on the Snowflake user and the client signs a token with the private key, common for service accounts.

Standard protocol Snowflake supports for automating user and role provisioning from an identity provider so account membership stays aligned with the corporate directory.

Native capability that lets a provider share database objects with consumer accounts without copying data; consumers query live, read-only data and pay for their own compute.

Account created and paid for by a provider so consumers without Snowflake can query shared data; readers can only query shares and cannot create their own objects.

Package of shared data or a Native App with metadata, terms, and pricing that a provider publishes to a private data exchange or to Snowflake Marketplace.

Public catalog where providers publish listings and consumers discover, request, and query third-party data and applications without copying them into their own accounts.

Copying databases, shares, and account objects from a source account to accounts in other regions or cloud platforms for disaster recovery, migration, or data residency.

Promoting a secondary replica to primary so applications resume with replicated objects; clients connect through a connection object that is repointed to the new primary account.

Snowflake-managed continuous ingestion that loads files from a stage shortly after they arrive, triggered by cloud event notifications or a REST call, using serverless compute billed per file.

API and SDK for writing rows directly into Snowflake tables at low latency without staging files, often used through the Kafka connector for streaming pipelines.

Named reference to a cloud storage location with URL, credentials, and encryption settings, used by COPY INTO and Snowpipe to read or write files.

Change data capture object that records inserts, updates, and deletes on a table, external table, or view and exposes them as rows with change metadata for downstream processing.

Scheduled or triggered unit of SQL, including a stored procedure call, used to run pipeline steps; tasks chain into DAGs where successors run after predecessors succeed.

Table whose contents Snowflake refreshes from a SQL definition according to a target lag, replacing hand-built scheduling, streams, and merge logic for incremental pipelines.

Table property that records which micro-partitions changed, allowing efficient incremental reads for streams and dynamic tables and enabling change queries without full table scans.

Account-level object tracking credit use by warehouses, with notify, suspend, or suspend-immediately actions at thresholds set on a schedule or a monthly quota.

Cost governance object that monitors actual and projected spend for the account or selected objects and sends alerts when spending crosses defined thresholds within a cycle.

Account usage view and interface listing executed queries with user, warehouse, duration, bytes scanned, and credits, used for tuning, workload analysis, and chargeback.

Billing unit consumed per second with a 60-second minimum while a warehouse runs, multiplied by its size factor; larger warehouses consume credits faster per hour of uptime.