3.5 Practice Drills and Readiness Markers

Resource Monitors and Credit Management

A resource monitor is an account-level object that tracks credit consumption by virtual warehouses and takes action when usage crosses a threshold. It controls compute credits only — it does not limit storage. By default only ACCOUNTADMIN can create resource monitors (the privilege can be delegated).

Key settings:

The three trigger actions are tested precisely:

• NOTIFY — send an alert only; warehouses keep running.
• SUSPEND — suspend assigned warehouses after running queries finish.
• SUSPEND_IMMEDIATE — suspend now, killing running queries.

A resource monitor can be assigned to the whole account or to specific warehouses (a warehouse can have only one monitor; the account can have one too). Trap: monitors only act on percentages of the credit quota and only suspend warehouses — they cannot stop storage growth or cloud-services usage.

Consumption-Based Billing

Snowflake bills on consumption, and the exam expects you to separate the cost buckets:

• Compute (virtual warehouse credits): billed per second (60-second minimum each time a warehouse resumes), scaled by warehouse size (XS=1 credit/hour, S=2, M=4, L=8, doubling per size).
• Storage: billed per TB per month based on average daily compressed storage, at a flat regional rate (no warehouse needed).
• Cloud Services / Serverless: the cloud-services layer is free up to 10% of daily compute credits; serverless features (Snowpipe, automatic clustering, materialized-view maintenance, replication, search optimization) consume their own credits.

Trap: suspending a warehouse stops compute charges but not storage charges — data still sits on disk. Likewise, Time Travel and Fail-safe storage add to storage cost. To control cost, you right-size and auto-suspend warehouses and use resource monitors; you cannot use a resource monitor to cap storage.

Worked example

A Medium (4 credits/hour) warehouse running 30 minutes consumes ~2 credits; with multi-cluster scaling to 3 clusters for that half hour it could consume ~6. Auto-suspend after 60 seconds of inactivity prevents idle credits — a frequent "how to reduce cost" answer.

Replication, Failover, and Readiness

Database replication keeps a secondary (read-only) copy of a database — or, with replication/failover groups, a set of account objects — in another Snowflake account, region, or cloud. Failover promotes that secondary to primary so clients can continue if the primary region is lost. Failover (the ability to promote) generally requires the Business Critical edition or higher. The basic flow: create a primary, replicate to a secondary, refresh on a schedule, and promote the secondary during a disaster. This underpins disaster recovery (DR) and business continuity.

Trap: replication is not the same as Secure Data Sharing (which gives live access without copying) and not the same as cloning (a zero-copy clone within one account). Replication physically copies data to another account for DR/migration.

Readiness markers

You are ready for this 20% domain when, for any stem, you can immediately:

1. Name the correct role (ORGADMIN/ACCOUNTADMIN/SECURITYADMIN/USERADMIN/SYSADMIN) and the least-privileged option that works.
2. List the grants needed to reach an object (USAGE chain + object privilege) and whether a future grant applies.
3. Pick the right authentication method (MFA, key-pair, SSO/SAML, OAuth) and apply a network policy correctly (blocked beats allowed).
4. Choose the right governance feature (masking vs. row access vs. tag-based masking vs. secure view) and the right audit view (ACCESS_HISTORY vs. INFORMATION_SCHEMA).
5. Apply the right cost control (resource monitor trigger, auto-suspend) and distinguish compute vs. storage charges.

If any marker is shaky, drill that specific topic rather than rereading the whole chapter.

Continuity and Cost Drills

Drill the continuity vocabulary until it is automatic, because three similar-sounding features are routinely confused:

Cost drill: for any "reduce spend" stem, the levers are auto-suspend and right-sizing warehouses, multi-cluster economy mode, resource monitors to cap and alert on credits, and pruning Time Travel retention to reduce storage. Remember a resource monitor only governs warehouse credits, with NOTIFY / SUSPEND / SUSPEND_IMMEDIATE triggers, and cannot cap storage.

Final self-check

Before moving on, run a quick oral quiz: name the six system roles in order; state the four grants needed to query a table; pick the auth method for a service account; choose the feature to mask all PII columns at once; and name the trigger that kills running queries. If you can rattle these off without hesitation, you have reached the readiness bar for a 20%-weighted domain. If you stumble on any one, that is the precise topic to revisit — targeted repair beats rereading, and it is the most efficient path to converting this domain's points on exam day.