Level 1 Overview

Level 1 lays the essential groundwork for everything that follows in this course. You'll begin by understanding exactly what ClickHouse is — an open source, column-oriented OLAP database built for real-time analytics at petabyte scale — and why it outperforms traditional data warehouses on speed, cost, and scalability. From there, you'll move into the architectural internals that make it possible to query billions of rows in just seconds: the MergeTree engine family, columnar storage, sparse primary indexes, granules, and partitioning. The level closes by demonstrating ClickHouse's versatility as a federated SQL query engine capable of reaching data wherever it lives, from object stores and data lakes to relational databases and local files.

By the end of Level 1, you'll have a thorough understanding of what ClickHouse is, how it stores and indexes data for maximum analytical performance, and how to query virtually any data source — all before writing a single line of transformation logic.

Module 1: What Is ClickHouse?

ClickHouse is an open source, column-oriented OLAP database management system purpose-built for real-time analytics, observability, machine learning and AI, and data warehousing. This module establishes the strategic case for ClickHouse: why it processes hundreds of millions to a billion rows per second, how it dramatically reduces storage costs through advanced compression, and why it integrates seamlessly with your existing data pipeline — from ingestion sources like ClickPipes, Fivetran, Apache Iceberg, and PostgreSQL to BI tools including Tableau, Looker, Power BI, and Grafana. ClickHouse is not a replacement for your transactional database; it is the high-performance analytical layer alongside it, and you pay only for the compute and compressed storage you actually use.

You'll also get a clear picture of the end-to-end ClickHouse Cloud workflow — bulk loading data via object stores like S3 using Parquet files, transforming it with materialized views that trigger automatically on insert, and surfacing insights through the visualization tools you already rely on. Real-world customer examples from companies like Disney+, Cloudflare, and Microsoft underscore that ClickHouse is production-proven at massive scale, delivering query times that are orders of magnitude faster than competing data warehouses at a lower and more predictable total cost of ownership.

What You'll Learn

• What ClickHouse is, where it came from, and what problems it was designed to solve
• The distinction between OLAP and OLTP workloads, and where ClickHouse excels
• Core use cases: real-time analytics, observability, machine learning and AI, and data warehousing
• How ClickHouse processes hundreds of millions to a billion rows per second
• How columnar storage eliminates unnecessary reads, greatly speeds up query times, and enables superior data compression
• Available integrations for ingestion (ClickPipes, S3, Fivetran, Iceberg, Postgres) How ClickHouse compares to and outperforms alternative data warehouses on speed and cost
• How to start a ClickHouse service and begin querying

Module 2: ClickHouse Architecture

This module uncovers the architectural foundations that make ClickHouse a uniquely powerful columnar database for analytical workloads. You'll learn how storing each column in a separate file drastically improves compression ratios and query timesAt the core of this architecture is the MergeTree table engine family — the default and recommended engine for all high-performance use cases. Data is written as immutable parts and automatically merged in the background, with the engine supporting hundreds of petabytes of storage while keeping query latency consistently low. You'll understand how each part stores its own primary index, and how background merges keep part counts healthy over time.

The module then goes deep on how cleverly choosing your primary key  can enable ClickHouseto skip enormous amounts of irrelevant data during query execution. Primary keys in ClickHouse determine sort order on disk, not uniqueness: the sparse primary index stores one entry per 8,192-row granule, and ClickHouse performs binary search over those marks to skip thousands of granules per query. You'll learn how to choose primary key columns strategically — ordering by ascending cardinality, selecting columns you filter on most frequently, and understanding the memory cost of each additional key. Partitioning with PARTITION BY is introduced as a tool for data lifecycle management and bulk deletion rather than query performance, with clear guidance on how high-cardinality partition keys can result in a ‘too many parts’ error.. Projections, secondary skipping indexes, and materialized views are introduced as additional optimization paths when a single primary key cannot serve all query patterns.

What You'll Learn

• How column-oriented storage differs from row-oriented storage and why it matters for real-time analytical performance
• The MergeTree table engine family
• What a granule is (8,192 rows by default) and how granules enable parallel query processing across CPU cores
• How the sparse primary index works and why it must fit entirely in memory
• Why primary keys in ClickHouse determine sort order rather than enforce uniqueness
• Best practices for primary key selection: ascending cardinality ordering, filter frequency, and memory cost trade-offs
• How partitioning organizes data into logical units for efficient data lifecycle management and deletion
• The dangers of high-cardinality partition keys and how to avoid "too many parts" errors

Module 3: ClickHouse as a Query Engine 

ClickHouse is not only a high-performance OLAP database — it is also a versatile SQL query engine capable of reaching data wherever it lives without requiring you to move it first. This module explores the full breadth of external sources ClickHouse can query directly: object stores including S3, Google Cloud Storage, Azure Blob Storage, and Cloudflare R2; open table formats like Apache Iceberg, Delta Lake, Apache Hudi, and Apache Paimon; and relational databases such as PostgreSQL, MySQL, and SQLite, as well as any system accessible via ODBC or JDBC. You'll see how this makes ClickHouse ideal for data exploration, ad hoc analysis, pre-load validation, and querying external systems you don't control — all using familiar SQL SELECT statements.

For recurring, performance-critical workloads, you'll understand the clear advantage of materializing data into ClickHouse natively, where the MergeTree engine, query cache, and materialized views combine to deliver consistently fast, sub-second analytical results. You'll also learn how federated queries allow you to join tables from entirely different data sources in a single SQL statement — a powerful capability for cross-system analysis without complex data pipelines. ClickHouse's extensive input/output format support (CSV, JSON, Parquet, and many more) and its REST interface make it the most flexible and accessible query engine available to both data engineers and web developers alike.

What You'll Learn

• How to use ClickHouse as a SQL query engine against data sources you don't need to migrate first
• Querying directly from S3, Google Cloud Storage, Azure Blob Storage, and Cloudflare R2
• Querying open table formats: Apache Iceberg, Delta Lake, Apache Hudi, and Apache Paimon
• Querying relational databases (PostgreSQL, MySQL, SQLite) via native connectors and ODBC/JDBC
• Supported input and output formats including CSV, JSON, Parquet, and many more
• How to write federated queries that join tables from multiple different data sources in a single SQL statement
• When to query data in-place versus when to materialize it into ClickHouse for recurring analytical workloads
• How the ClickHouse query cache and materialized views accelerate frequently executed analytical queries
• How to import local files directly through the ClickHouse Cloud UI