Read every Teach first section before you copy SQL.
This lab is the exam’s architecture foundation: where data lives, what burns credits,
and which edition unlocks which feature. Domain 1 tips after you finish.
Business problem
Morgan asks: “If we resize the warehouse, do we get more storage?”
Kai thinks Snowflake is one blob. Wrong answers here wipe ~⅓ of Core.
Solution you will build
Name the three layers out loud, map editions (Standard → VPS) to Time Travel / MCW / Tri-Secret / PrivateLink,
and run tiny SQL that proves compute ≠ storage (plus permanent vs transient).
Before
One blurry “Snowflake”
“Warehouse stores our tables”
No idea which edition we need
Cloud Services = mystery fee
After
Three layers + edition map
Storage · Compute · Cloud Services
90-day TT → Enterprise+
WH = engine; tables stay in storage
Episode 01 · Start here
Kai mixes up the engine and the shelves.
KaiAnalyst — still learning
Morgan said scale the warehouse for the board deck. So… bigger warehouse = more room for history tables, right?
DevCoach — explains why
No. Warehouse = kitchen staff (compute). Tables live in the storage pantry. Cloud Services is the front desk (login, catalog, planner). Three layers — say them every time.
Exam trap #1 lives here: never say “the warehouse stores data.”
See it — three layers with Alpine names
Layer 1
Database Storage
Compressed micro-partitions on cloud object storage. Your tables live here.
ALPINE_RETAIL.LAB.CUSTOMERS bytes → storage bill
Layer 2
Query Processing
Virtual warehouses run SQL. Credits while running / resumed.
Standard tops out at 1 day retention. Exam loves this distractor.
“We need 90-day UNDROP” → Enterprise (or higher)
Security extras
Tri-Secret & PrivateLink
Customer-managed keys + private connectivity → Business Critical (and VPS).
Sam: “keys we control” → Business Critical+
Before Step 1
Morgan: “Where are we running?”
MorganManager — deadline
If compliance asks which cloud region Alpine’s account sits in, can you answer in ten seconds?
DevCoach
CURRENT_ACCOUNT() and CURRENT_REGION(). Cloud Services knows who you are; storage holds the bytes in that region’s cloud.
1 Prove who / where you are
What this does: Reads session identity and account location — Cloud Services metadata, not a warehouse scan of table bytes.
Why: Exam questions mix “account locator,” region, and warehouse. Separate them.
SQL · account context
USE ROLE ROLE_CORE_LAB_ADMIN;
USE WAREHOUSE WH_CORE_XS;
USE DATABASE ALPINE_RETAIL;
USE SCHEMA LAB;
SELECT CURRENT_ACCOUNT() AS account_locator,
CURRENT_REGION() AS region,
CURRENT_ROLE() AS role,
CURRENT_WAREHOUSE() AS warehouse;
✓ You should see a locator, a region string, ROLE_CORE_LAB_ADMIN, WH_CORE_XS
Before Step 2
Kai: “What are PARAMETERS?”
KaiAnalyst
Is DATA_RETENTION_TIME_IN_DAYS a warehouse setting? Feels like compute…
DevCoach
Retention is about storage history (Time Travel window). Edition caps how high you can set it. Warehouse size won’t give you 90 days.
2 Peek at retention parameters
What this does: Shows account-level knobs tied to Time Travel retention.
What you should see: A value (often 1 on trials). Remember: Enterprise unlocks up to 90 days — your trial number ≠ the exam rule.
SQL · SHOW PARAMETERS
SHOW PARAMETERS LIKE 'DATA_RETENTION_TIME_IN_DAYS'
IN ACCOUNT;
SHOW PARAMETERS LIKE 'MIN_DATA_RETENTION_TIME_IN_DAYS'
IN ACCOUNT;
Do we need Fail-safe on every scratch table Priya builds for demos?
DevCoach
No. Transient skips Fail-safe (cheaper). Permanent keeps Fail-safe after Time Travel. Lab 04 digs deeper — today just feel the difference.
3 Permanent vs transient tiny demo
What this does: Creates one permanent and one transient table so you can say “Fail-safe: yes vs no.”
SQL · permanent vs transient
CREATE OR REPLACE TABLE LAB.DEMO_PERM_TINY (
ID NUMBER, NOTE STRING
);
CREATE OR REPLACE TRANSIENT TABLE LAB.DEMO_TRANS_TINY (
ID NUMBER, NOTE STRING
);
INSERT INTO LAB.DEMO_PERM_TINY VALUES (1, 'has Fail-safe after TT');
INSERT INTO LAB.DEMO_TRANS_TINY VALUES (1, 'no Fail-safe');
SHOW TABLES LIKE 'DEMO_%' IN SCHEMA LAB;
✓ Two tables listed — note TRANSIENT kind on one of them
4 Say it: warehouse is not storage
What this does: Reminds you the session warehouse is only the compute handle.
SQL · compute context
SELECT CURRENT_WAREHOUSE() AS compute_engine_only;
-- Layers (exam — not SQL objects):
-- Storage | Query Processing (WH) | Cloud Services
✓ Returns WH_CORE_XS — say aloud: “engine, not pantry”
Exam presses further — say these out loud
Cover each answer before you open it. Domain 1 traps.
Q1. Does a virtual warehouse store table data?
Say: No. Warehouses are Query Processing (compute). Table data lives in Database Storage as compressed micro-partitions. Bigger warehouse ≠ more storage.
Q2. Which edition do you need for up to 90-day Time Travel?
Say: Enterprise or higher. Standard is capped at 1 day. Don’t confuse “my trial shows X” with the edition matrix.
Q3. How is Cloud Services billed?
Say: Cloud Services usage is metered; Snowflake applies a daily credit against warehouse compute. Heavy metadata / services beyond that can show as Cloud Services charges — it is not “free forever,” and it is not the same as warehouse credits.
Q4. What’s the benefit of separating storage and compute?
Say: Scale and pay for each independently: keep large history in storage while using a small warehouse for light queries, or spin up larger compute for spikes without moving data.
Q5. Tri-Secret Secure and AWS PrivateLink — which edition tier?
Say: Business Critical (and VPS). Enterprise adds Multi-cluster warehouses and longer Time Travel, but those hardened networking / key features are Business Critical+.
Q6. Name the three architecture layers in exam order.
Priya’s wrap-up (say this out loud)
“Alpine’s Snowflake is three layers: storage holds our micro-partitions, warehouses like WH_CORE_XS are compute only,
and Cloud Services handles login, metadata, and planning. Editions matter — 90-day Time Travel needs Enterprise+;
Tri-Secret and PrivateLink need Business Critical+. Transient tables skip Fail-safe. I will never say the warehouse stores data.”
Next
Lab 02 walks the object catalog — tables, views, stages, sequences, and what’s account-level.