COF-C03 domain - 31% of the exam

Snowflake AI Data Cloud Features and Architecture

Snowflake AI Data Cloud Features and Architecture is 31% of the COF-C03 exam. These are the objectives it covers, each with practice questions, with every answer explained.

The domain in numbers

86
Practice questions
8
Objectives
20
Easy
66
Medium

Where people go wrong

  • Scaling up to clear concurrency queuing

    Scaling up is offered as the cure for queuing. A larger warehouse size speeds each query and adds memory, but the queries still run on one cluster with a limited number of concurrency slots, so a burst of short dashboard queries keeps waiting. Doubling the size is not the primary lever for concurrency. Queuing under concurrency is cleared by scaling out: a multi-cluster warehouse in auto-scale mode starts extra clusters when queries queue and stops them when demand falls.

    See it in a sample question
  • Assigning a feature to the wrong edition

    The wrong answers place a feature at the wrong edition. Standard is offered for Tri-Secret Secure, private connectivity, materialized views or multi-cluster warehouses, and Enterprise is offered for database failover and failback. Business Critical and Virtual Private Snowflake appear as the answer when a lower edition already has the feature. Multi-cluster warehouses, materialized views and extended Time Travel start at Enterprise. Tri-Secret Secure, private connectivity, and failover and failback start at Business Critical.

    See it in a sample question
  • Billing compute and storage on the wrong measure

    The wrong answers move billing onto the wrong measure. They charge compute on bytes scanned, user count or micro-partitions, and storage in warehouse credits or on peak or uncompressed size. Warehouse credits follow warehouse size and running time. Storage is a separate dollar charge per terabyte on the average daily compressed data, at a discounted rate for pre-purchased capacity storage or a higher on-demand rate. Edition changes the dollar value of a credit, not the credits consumed.

    See it in a sample question
  • Confusing Snowsight, SnowSQL, the CLI and drivers

    The wrong answers hand a requirement to the neighbouring client. Snowsight is the browser interface for worksheets, dashboards and charts, and is offered for scripted terminal runs. SnowSQL is the older command-line client for running SQL and loading files, and is offered where the modern tool is needed. The Snowflake CLI uses connection profiles to manage objects and deploy Streamlit and Snowpark apps from CI. Drivers and connectors are libraries an application calls, never a worksheet or command-line tool.

    See it in a sample question
  • Expecting a standard or materialized view to hide its definition

    The wrong answers offer a standard view or a materialized view where the requirement is to hide the SQL definition and base table names. A standard view exposes its definition to roles that can describe it, and a materialized view exists for performance, so its defining query stays inspectable. Only a secure view conceals the definition from consumers and bypasses optimiser pushdowns that could leak base data, which is why it suits a share.

    See it in a sample question

Objectives in this domain

What to study

The official documents this domain's practice questions cite most.

Sample questions from this domain

Free sampleSnowflake AI Data Cloud Features and Architecturemedium

During business hours hundreds of short dashboard queries hit one warehouse at once, and many of them queue and wait because the single cluster is saturated, even though each query is small. The team wants queries to stop queuing during these concurrency spikes without making individual queries run on larger compute. Which configuration best addresses this?

  • AConfigure the warehouse as a multi-cluster warehouse in auto-scale mode so Snowflake starts extra clusters when queries queue and stops them when demand falls Correct
  • BIncrease the warehouse to a larger size so each dashboard query gets more memory and the extra capacity clears the queue during the busy period
  • CSet a long statement timeout on the warehouse so queued queries are allowed to wait longer and eventually all run on the existing single cluster
  • DLower the auto-suspend setting to zero so the warehouse never suspends and is therefore always warm enough to absorb the concurrency spikes by itself
Identify a multi-cluster warehouse in auto-scale mode as the way to clear query queuing caused by high concurrency. Query queuing under high concurrency is a scale-out problem because a single cluster has a finite number of concurrent execution slots. A multi-cluster warehouse in auto-scale mode starts additional clusters when Snowflake detects queuing and shuts them down when load drops, spreading concurrent queries across clusters rather than enlarging each query's compute.

Why A is correct: A multi-cluster warehouse in auto-scale mode adds clusters when queuing is detected, spreading the concurrent dashboard queries across clusters so they stop waiting, then scales back down.

Why B is wrong: A larger size speeds up each query but still runs them on one cluster with limited concurrency slots, so queries keep queuing under a heavy concurrent load.

Why C is wrong: A longer statement timeout only lets queries wait before being cancelled, so it does not add concurrency and the queue still builds during the spike.

Why D is wrong: Keeping a single-cluster warehouse always running does not add concurrency capacity, so queries still queue when the one cluster is saturated during the spike.

Free sampleSnowflake AI Data Cloud Features and Architecturemedium

A healthcare analytics team must store regulated patient data in Snowflake and has a contractual requirement for customer-managed encryption keys (Tri-Secret Secure) and support for private connectivity. Which is the lowest Snowflake edition that satisfies both requirements?

  • AStandard edition, because it already includes always-on encryption of data at rest and in transit for every account.
  • BEnterprise edition, because it raises Time Travel retention and adds materialized views for sensitive workloads.
  • CBusiness Critical edition, because it adds Tri-Secret Secure customer-managed keys and supports private connectivity to the Snowflake service. Correct
  • DVirtual Private Snowflake, because a completely isolated environment is mandatory before customer-managed keys can be used at all.
Identify Business Critical as the lowest Snowflake edition that adds Tri-Secret Secure and private connectivity for regulated data. Customer-managed keys through Tri-Secret Secure and private connectivity options are introduced at the Business Critical edition. Standard and Enterprise lack them, while Virtual Private Snowflake is a higher and more isolated edition, so Business Critical is the lowest edition that meets both stated needs.

Why A is wrong: Standard does encrypt data at rest and in transit, but it does not offer Tri-Secret Secure or private connectivity, so it cannot meet the stated requirement.

Why B is wrong: Enterprise adds extended Time Travel and materialized views, but Tri-Secret Secure and private connectivity only appear at Business Critical, so it is insufficient.

Why C is correct: Business Critical is the lowest edition that adds Tri-Secret Secure and private connectivity, which together satisfy the regulated-data requirement stated.

Why D is wrong: Virtual Private Snowflake also includes these features but is a higher, isolated edition, so it is not the lowest edition that meets the requirement.

Free sampleSnowflake AI Data Cloud Features and Architecturemedium

A finance lead is comparing two ways Snowflake can price the same stored data and asks how pre-purchased capacity storage differs from on-demand storage. Which statement correctly describes the difference between the two storage pricing options?

  • AOn-demand storage is always free up to a fixed monthly allowance, while capacity storage charges a flat fee that ignores how much data is actually held.
  • BCapacity storage is a discounted rate from committing to a pre-purchased amount up front, while on-demand storage charges a higher per-terabyte rate billed in arrears. Correct
  • CCapacity storage bills per credit consumed by warehouses, whereas on-demand storage bills a separate dollar amount per terabyte of data held.
  • DOn-demand storage measures uncompressed bytes and capacity storage measures compressed bytes, so the two options bill different physical data volumes.
Distinguish pre-purchased capacity storage pricing, a discounted upfront commitment, from on-demand storage billed in arrears at a higher per-terabyte rate. Both storage options charge in dollars per terabyte on the average daily compressed data held, so they meter the same physical volume. The difference is commercial: capacity storage offers a lower per-terabyte rate in return for committing to a pre-purchased amount, while on-demand storage charges a higher rate and bills actual usage in arrears without any commitment. The rate and commitment model, not the measurement basis, is what separates them.

Why A is wrong: On-demand storage is not free up to an allowance, and capacity storage still meters actual usage; both bill for the average data stored, so this contrast is wrong.

Why B is correct: Capacity storage gives a lower rate in exchange for an upfront commitment, while on-demand bills usage after the fact at a higher per-terabyte rate, which is the correct distinction.

Why C is wrong: Storage of either kind is billed in dollars per terabyte, not in warehouse credits, so describing capacity storage as credit-based confuses storage with compute.

Why D is wrong: Both options meter the same compressed data volume; the pricing difference is the rate and commitment model, not whether bytes are compressed, so this is incorrect.

Free sampleSnowflake AI Data Cloud Features and Architectureeasy

A platform engineer wants a modern command-line developer tool that uses connection profiles to manage Snowflake objects and deploy applications such as Streamlit and Snowpark from scripts and CI pipelines. Which tool fits this purpose?

  • AThe Snowflake CLI, because it is a developer command-line tool that uses connection profiles to manage objects and deploy apps like Streamlit and Snowpark from scripts and CI Correct
  • BSnowSQL, because it is the older command-line client for running ad hoc queries and scripts, so it deploys Streamlit and Snowpark apps through connection profiles from CI pipelines
  • CA Snowflake Python connector, because it is a code library for opening sessions, so it acts as a CLI that manages objects and deploys Streamlit and Snowpark apps from CI pipelines
  • DSnowsight worksheets, because they run in the browser, so they serve as a command-line developer tool with connection profiles to deploy Streamlit and Snowpark apps from CI pipelines
Identify the Snowflake CLI as the command-line developer tool for object management and app deployment via connection profiles. The Snowflake CLI is a command-line developer tool that uses named connection profiles to manage Snowflake objects and deploy applications such as Streamlit and Snowpark, which makes it well suited to scripted and CI-driven workflows.

Why A is correct: The Snowflake CLI is the modern command-line developer tool that uses named connections to manage objects and deploy Streamlit and Snowpark applications, including from CI pipelines, which fits exactly.

Why B is wrong: SnowSQL is the legacy command-line query client; it executes SQL but does not use the modern connection-profile model to manage objects or deploy Streamlit and Snowpark applications.

Why C is wrong: A connector or driver is a library you import into application code to open connections, not a command-line tool, so it does not manage objects or deploy apps on its own from CI.

Why D is wrong: Snowsight worksheets are a browser interface for writing and running SQL, not a command-line tool, so they cannot use connection profiles or drive scripted CI deployments.

Free sampleSnowflake AI Data Cloud Features and Architecturemedium

A data team shares a complex join across several base tables with downstream analysts. They want consumers to query it like a table, but they must hide the underlying SQL definition and the names of the base tables from anyone who inspects the object. Which object meets this requirement?

  • AA standard view, because its definition is computed at query time and Snowflake hides the underlying SQL text from every consumer by default
  • BA materialized view, because it stores precomputed results and therefore conceals the base table names and the query that produced those results
  • CA transient table loaded by a scheduled task, because copying the join output into a table removes any link back to the original base tables
  • DA secure view, because it both behaves like a table at query time and prevents consumers from seeing the view definition and the underlying tables Correct
Select a secure view when the view must behave like a table while hiding its definition and underlying base objects from consumers. A secure view enforces that the view definition and the identity of the underlying objects are not exposed to consumers and bypasses certain query optimisations that could leak base data, so it is the object that combines table-like access with definition hiding.

Why A is wrong: A standard view does run at query time, but its definition is visible to roles that can describe it, so it does not hide the underlying SQL as required.

Why B is wrong: A materialized view precomputes results for speed, but it does not hide its definition; consumers can still inspect the defining query and base tables.

Why C is wrong: Copying into a transient table loses the live view behaviour and still does not formally protect a definition, so it does not satisfy querying it like a managed view.

Why D is correct: A secure view is queried like a table yet hides its SQL definition and base objects from consumers, which is exactly the concealment the team needs.

Other domains in this exam

See also the COF-C03 cert hub, the study guide, and the cheat sheet.

Examworthy is not affiliated with or endorsed by Snowflake. Original, blueprint-aligned practice material only.