Snowflake Tutorials

Snowflake from a first query to a warehouse you can put a product on — the architecture that makes storage and compute separable, sizing virtual warehouses, loading data with stages and Snowpipe, querying JSON without a schema, making slow queries fast, Time Travel, streams and tasks, role-based access control, and keeping the credit bill predictable. Every query runs against SNOWFLAKE_SAMPLE_DATA, which your account already has on day one.

• Snowflake – Putting It in Production
  The checklist between a working warehouse and one other people depend on. Separating dev, staging and production, versioning DDL in git and deploying it in CI, key-pair auth and network policies for service accounts, monitoring and alerting on the views that matter, and the failure modes worth having an answer for before they happen.
• Snowflake – Understanding and Controlling Cost
  Where the credits actually go, and the handful of changes that move the bill most. Reading ACCOUNT_USAGE for warehouse, storage and serverless spend, resource monitors that suspend rather than warn, the auto-suspend and idle-warehouse waste everybody has, and why the fastest query is usually also the cheapest one.
• Snowflake – Roles, Grants and Access Control
  Snowflake's RBAC in the order you need to understand it. Why ACCOUNTADMIN is not your working role, ownership versus grants, future grants — the single feature that stops permissions rotting every time someone adds a table — and a functional role layout you can copy. Plus masking policies and row access policies, briefly.
• Snowflake – Streams and Tasks
  Native change tracking and scheduling, so a pipeline needs no extra orchestrator. What a stream really is (an offset, not a copy), the consumption rule that surprises everyone, task trees and serverless tasks, and the standard pattern: a stream feeds a MERGE, a task runs it only when there is something to do.
• Snowflake – Time Travel, Cloning and Undrop
  The two features that change how you work, not just what you can recover. Querying a table as it was before the bad UPDATE, UNDROP, zero-copy clones that give you a full copy of production for a dev branch at no storage cost, and Fail-safe — which is Snowflake's insurance policy, not your backup.
• Snowflake – Making Queries Fast
  Why a query is slow, in the order worth checking. Reading the Query Profile, pruning and what breaks it, spilling to local and remote storage and what to do about it, the three caches and which one you are actually hitting, and when clustering keys earn their cost — which is far less often than people reach for them.
• Snowflake – Querying Data
  The SQL that is worth knowing beyond SELECT ... WHERE. CTEs, window functions, QUALIFY — the clause Snowflake has and Postgres does not — SAMPLE for exploring a large table cheaply, GROUP BY ALL, and the date and string functions that come up in every reporting query. All of it against TPCH_SF1.
• Snowflake – JSON and Semi-Structured Data
  Loading JSON without designing a schema first, then querying it as if you had. The colon path syntax, casting out of VARIANT and why forgetting to cast makes strings come back with quotes, FLATTEN for arrays, and the decision that actually matters: when to leave data in a VARIANT column and when to shred it into real columns.
• Snowflake – Snowpipe and Continuous Loading
  When a nightly COPY is not enough. Snowpipe with cloud-storage event notifications, how its serverless billing differs from a warehouse, monitoring a pipe that has quietly stopped, and where the newer options — Snowpipe Streaming and directory tables — fit. Includes the notification setup that is the only genuinely fiddly part.
• Snowflake – Loading Data with Stages and COPY INTO
  Getting files in. Internal versus external stages, PUT and COPY INTO, file formats as reusable objects, why a hundred medium files load faster than one huge one, and what to do when a row fails — ON_ERROR, VALIDATION_MODE and the load metadata that silently skips a file you already loaded.
• Snowflake – Data Types That Matter
  The short list worth knowing, and the three that cause real bugs. Why NUMBER is the only sensible choice for money, why VARCHAR(16777216) costs nothing extra, the three TIMESTAMP flavours and which one to standardise on, and VARIANT — the type that lets a column hold JSON and still be queried with SQL.
• Snowflake – Databases, Schemas and Tables
  The object hierarchy and the DDL you will use every day. Permanent, transient and temporary tables and what the difference costs you, views versus materialised views, why Snowflake accepts a primary key and then ignores it, CREATE TABLE ... LIKE and CLONE, and the naming and identifier-casing rules that bite everyone exactly once.
• Snowflake – Virtual Warehouses and Sizing
  The compute you are actually paying for. What a size means in credits, why doubling the size can cost the same as leaving it alone, auto-suspend and auto-resume and the one setting people get wrong, multi-cluster warehouses for concurrency versus a bigger warehouse for one slow query, and how to split workloads so a loading job cannot slow down a dashboard.
• Snowflake – Getting Started: Account, Worksheet and Connectors
  From a trial signup to your first query against SNOWFLAKE_SAMPLE_DATA. Picking an edition and a region, the four things every session needs set, running SQL in Snowsight, and connecting from outside the browser — the Snowflake CLI, the Python connector and JDBC — with the key-pair auth you should use instead of a password.
• Snowflake – Architecture: Storage, Compute and Cloud Services
  The three layers, and why every cost and performance decision later in this track comes back to them. Micro-partitions and how they replace the indexes you are used to, why two teams can query the same table at full speed without blocking each other, and what the services layer does for free — including the result cache that makes a repeated query cost nothing at all.
• Snowflake – What It Is and Why It Exists
  Start here. What Snowflake actually is — a SQL data warehouse you rent by the second, on somebody else's cloud — and what problem it was built to solve that Postgres and Hadoop did not. What you get on day one, what you pay for, when it is the wrong tool, and the lesson index for this track in reading order.