Skip to content
Tutorials Center
My learning

Topics

Lesson 1 of 38 Β· Getting Started

What Is Snowflake?

Learn what Snowflake is, what problems it solves, how it differs from a classic database, and how this course runs its examples without a Snowflake account.

Snowflake is a cloud data platform. You give it data, and it stores that data, runs SQL over it and shares the results, while you never install a server, tune a disk or patch an operating system. It runs on Amazon Web Services, Microsoft Azure and Google Cloud, and the same SQL works on all three.

Most people meet Snowflake as a cloud data warehouse: a system built for analytics, where questions touch millions or billions of rows and answers feed dashboards and reports. Over the years it has grown into a platform that also loads streams of events, runs Python next to the data, shares tables with other companies and serves data applications. This chapter gives you the map. The rest of the course walks through each region.

How the course runs its examples

A note on honesty before we start. Snowflake is a commercial cloud service, and this website cannot connect to one. So this course uses two kinds of code:

  • Runnable examples use standard SQL in a small in-browser database. They teach ideas that carry straight over to Snowflake: joins, grouping, window functions, JSON handling. When Snowflake spells something differently, the lesson says so and shows the Snowflake version next to it.
  • Read-only examples show Snowflake-only syntax such as CREATE WAREHOUSE or COPY INTO. These blocks are labeled Snowflake SQL (read-only). The output you see described is what Snowflake returns, but you cannot run them here.

Where an in-browser example stands in for a Snowflake feature, it is labeled stand-in. You will always know which is which.

Snowflake offers a free 30-day trial with credits, so you can type every read-only example into a real account later. Accounts, Editions and the Web UI explains how.

The problem Snowflake solves

Classic data warehouses tied storage and computing power together. If you needed more speed you bought a bigger machine, and that machine also held your data, so you paid for capacity even at 3 a.m. when nobody queried anything. Different teams fought over the same CPUs. Loading data at night slowed the morning reports.

Snowflake separates the pieces:

  • Data lives once, in cheap cloud storage, in a compressed columnar format.
  • Virtual warehouses (clusters of compute) read that data. You can start as many as you like, size each one independently and pause them when idle.
  • A cloud services layer handles security, metadata, query planning and transactions.

Because each team can have its own warehouse reading the same tables, the finance dashboard no longer slows down the data science notebook. You will see exactly how in The Three-Layer Architecture.

What you get

Fully managed No servers to run
Elastic Scale compute up and down
One copy of data Many teams read it
Plain SQL ANSI-style language

You pay for storage and for the compute you actually use.

1 / 2

What makes it different from a classic database

If you know MySQL or PostgreSQL, a few things will surprise you:

Idea Classic OLTP database Snowflake
Main job Many small reads and writes (an app) Few huge scans and aggregations (analytics)
Indexes You create and tune them None to create; data is organized in micro-partitions
Storage and compute Same server Separate, each scaled alone
Tuning Indexes, vacuum, memory settings Warehouse size and data layout
Billing Server hours Credits per second of compute plus storage

Snowflake is not the right tool for a shopping-cart application that updates one row a thousand times per second. It is excellent when a question starts with "across all of last year's orders…".

Standard SQL, a few extras

The language is SQL. Here is a query you could run on any database, including the in-browser one below. It counts books by genre and shows the average price:

SQL A first analytic query Β· database: bookstore

Press Run (or Ctrl + Enter) to see the result here.

In Snowflake the statement is character-for-character the same. What differs is where it runs and how fast it scans a table with a billion rows. Snowflake adds extras on top of standard SQL, such as QUALIFY for filtering window functions, a VARIANT type for JSON, and FLATTEN for unpacking arrays. You will meet all of them in this chapter and in Querying and Semi-structured Data.

The objects you will work with

A quick preview of the vocabulary used all through the course:

  • Account: your Snowflake environment, with its own users, data and bill.
  • Database and schema: folders for tables, in two levels.
  • Table, view: where data lives and how it is presented.
  • Warehouse: compute that runs queries.
  • Stage: a place for files waiting to be loaded or unloaded.
  • Role: a bundle of privileges that users act under.

Here is the same idea written as read-only Snowflake SQL. Do not worry about every keyword yet:

SQL
-- Snowflake SQL (read-only)
CREATE WAREHOUSE learn_wh WAREHOUSE_SIZE = 'XSMALL' AUTO_SUSPEND = 60;
CREATE DATABASE bookshop;
CREATE SCHEMA bookshop.sales;
USE WAREHOUSE learn_wh;
USE SCHEMA bookshop.sales;
SELECT CURRENT_WAREHOUSE(), CURRENT_DATABASE(), CURRENT_SCHEMA();

Four statements create the compute and the folders, and the last one asks Snowflake what you are currently using. Try it on a trial account later.

A tiny stand-in for "many teams, one copy"

The headline idea is that several consumers read the same data. In the stand-in below, one table is read through two different views. Think of each view as a different team's perspective on the single stored copy. This is only an analogy: in Snowflake the two teams would also use two separate warehouses.

SQL Stand-in: one table, two perspectives Β· database: bookstore

Press Run (or Ctrl + Enter) to see the result here.

Check your understanding

Quick check Β· 4 questions

of 4 answered You scored / 4 Β· perfect!

  1. 1. What is Snowflake's main design idea compared with a classic data warehouse?

    Data sits once in cloud storage while virtual warehouses provide compute. Because they are separate, you can resize or pause compute without moving data.
  2. 2. Which workload fits Snowflake best?

    Snowflake is built for analytics: large scans, joins and aggregations. High-rate single-row transaction workloads are better served by an OLTP database.
  3. 3. In this course, what does a block labeled 'Snowflake SQL (read-only)' mean?

    The site cannot connect to a Snowflake account, so product-specific statements are shown as read-only. Portable SQL examples run in the browser.
  4. 4. Do you create indexes in Snowflake to speed up queries?

    Standard tables have no user-created indexes. Snowflake splits tables into micro-partitions and prunes them using metadata.

Summary

  • Snowflake is a fully managed cloud data platform, most often used as a data warehouse.
  • Storage, compute and cloud services are separate layers, so you scale and pay for each independently.
  • The query language is SQL with useful extras like QUALIFY, VARIANT and FLATTEN.
  • It targets analytics, not high-rate single-row transactions.
  • In this course, portable SQL runs in the browser and Snowflake-only syntax is read-only and clearly labeled.

Next: Accounts, Editions and the Web UI