Creating and Using Materialized Views in ClickHouse (Performance Optimization)

Materialized views are an important feature in ClickHouse used to improve query performance.

What is a Materialized View?

A materialized view is a pre-calculated view whose internal data is stored locally on disk, allowing it to be read directly from the disk during each query without needing complex calculations.

How to Create a Materialized View?

Step 1: Select the Table to Materialize

In ClickHouse, all usable tables can serve as the basis for materialized views. For example:

CREATE TABLE my_table (
  id Int64,
  name String(255),
  age Int64
) ENGINE = MergeTree;

Step 2: Define the Materialized View

Create a materialized view using the CREATE VIEW statement:

CREATE VIEW materialized_view AS
SELECT *
FROM my_table
WHERE age > 18;

Using Materialized Views

Advantages

Materialized views have several significant advantages, including:

• Performance Improvement: Materialized views can read data directly from the disk, thereby reducing computation time and improving query performance.
• Reliability: Since pre-calculated data is stored locally on the disk, materialized views provide higher reliability and availability.

Use Cases

Here are some typical use cases:

• Reports and Statistics: For reports that need frequent updates but do not require reflecting the latest data in real-time, materialized views can be used to improve performance.
• Data Analysis: For large datasets and complex calculations, materialized views can quickly retrieve pre-calculated results.

Performance Optimization

1. Set Appropriate Filter Conditions

In a materialized view, filter conditions reduce the number of rows read from the original table, thereby improving performance.

CREATE VIEW materialized_view AS
SELECT *
FROM my_table
WHERE age > 18 AND country = 'USA';

2. Use Appropriate Sorting Columns

Use sorting columns to quickly find the data needed in a materialized view.

CREATE VIEW materialized_view AS
SELECT id, name
FROM (
  SELECT id, name, row_number() OVER (ORDER BY age) as row_num
  FROM my_table
) sub
WHERE row_num > 10;

3. Use Suitable Storage Engines

Choose a storage engine that is suitable for the size and type of materialized view data. MergeTree, HashJoin, and other engines are generally applicable in most cases.

CREATE TABLE my_table (
  id Int64,
  name String(255),
  age Int64
) ENGINE = MergeTree;

4. Reduce Filter Conditions

Reducing filter conditions can improve the performance of materialized views.

CREATE VIEW materialized_view AS
SELECT *
FROM my_table;

Note: Increasing filter conditions will reduce the availability of a materialized view.