Clustering keys for large tables in Snowflake - Commands & Configuration

Introduction

Large tables can become slow to query because data is not organized efficiently. Clustering keys help Snowflake organize data in a way that speeds up queries by grouping related rows together.

When you have a very large table and queries filter on specific columns often

When you notice queries scanning too much data and running slowly

When you want to improve performance without manually partitioning data

When your table grows continuously and you want to keep data organized automatically

When you want to reduce the cost of querying large datasets by scanning less data

Commands

This command adds a clustering key on the columns 'region' and 'sale_date' to the 'sales_data' table. It tells Snowflake to organize data based on these columns to speed up queries filtering on them.

Terminal

ALTER TABLE sales_data CLUSTER BY (region, sale_date);

Expected Output

ALTER TABLE

This command checks how well the 'sales_data' table is clustered. It shows if the clustering key is effective or if Snowflake needs to reorganize data.

Terminal

SELECT SYSTEM$CLUSTERING_INFORMATION('sales_data');

Expected Output

{

  "clustering_depth": 3,

  "average_depth": 2.5,

  "partition_count": 10

}

This command tells Snowflake to reorganize the data in 'sales_data' based on the clustering key. It helps improve query performance by physically sorting the data.

Terminal

ALTER TABLE sales_data RECLUSTER;

Expected Output

ALTER TABLE

This query benefits from the clustering key by scanning less data when filtering on 'region' and 'sale_date'. It runs faster because data is organized by these columns.

Terminal

SELECT COUNT(*) FROM sales_data WHERE region = 'East' AND sale_date = '2023-05-01';

Expected Output

12345

Key Concept

If you remember nothing else from this pattern, remember: clustering keys help Snowflake organize large tables to speed up queries filtering on specific columns.

Common Mistakes

Adding clustering keys on columns that are rarely used in filters

Snowflake spends resources organizing data that does not improve query speed, wasting compute and storage.

Choose clustering keys based on columns frequently used in WHERE clauses or join conditions.

Not running RECLUSTER after adding or changing clustering keys

Data remains unordered physically, so queries do not benefit from clustering until reclustering happens.

Run ALTER TABLE RECLUSTER to reorganize data after setting clustering keys.

Expecting immediate query speedup without checking clustering effectiveness

Clustering may not be effective if data is not well distributed or if clustering depth is low.

Use SYSTEM$CLUSTERING_INFORMATION to monitor clustering quality and adjust keys if needed.

Summary

Use ALTER TABLE CLUSTER BY to set clustering keys on columns used often in queries.

Check clustering effectiveness with SYSTEM$CLUSTERING_INFORMATION.

Run ALTER TABLE RECLUSTER to physically reorganize data and improve query speed.

Practice

1.

What is the main purpose of using clustering keys in large Snowflake tables?

easy

A. To group data physically for faster query performance

B. To encrypt the data for security

C. To create backups automatically

D. To increase storage space usage

Solution

1. Step 1: Understand clustering keys role

   Clustering keys organize data physically in storage to improve how fast queries run.

2. Step 2: Compare options

   Only To group data physically for faster query performance describes grouping data for faster queries. Others mention unrelated features like encryption or backups.

3. Final Answer:

   To group data physically for faster query performance -> Option A

4. Quick Check:

   Clustering keys = faster queries [OK]

Hint: Clustering keys speed up queries by grouping data [OK]

Common Mistakes:

• Thinking clustering keys encrypt data
• Confusing clustering keys with backups
• Assuming clustering keys increase storage

2.

Which of the following is the correct syntax to add a clustering key on column user_id when creating a table in Snowflake?

CREATE TABLE users (id INT, user_id INT, name STRING) ...

easy

A. CLUSTERING user_id

B. CLUSTER ON user_id

C. CLUSTER KEY user_id

D. CLUSTER BY (user_id)

Solution

1. Step 1: Recall Snowflake clustering syntax

   Snowflake uses CLUSTER BY (column_name) to define clustering keys during table creation.

2. Step 2: Check options

   Only CLUSTER BY (user_id) matches the correct syntax. Others use incorrect keywords or missing parentheses.

3. Final Answer:

   CLUSTER BY (user_id) -> Option D

4. Quick Check:

   Correct syntax = CLUSTER BY (column) [OK]

Hint: Use 'CLUSTER BY (column)' to set clustering keys [OK]

Common Mistakes:

• Using 'CLUSTER ON' instead of 'CLUSTER BY'
• Omitting parentheses around column name
• Using 'CLUSTER KEY' which is invalid

3.

Given the table sales with columns region, date, and amount, and clustering key on region, what is the expected effect on the query below?

SELECT SUM(amount) FROM sales WHERE region = 'West';

medium

A. Query result will be incorrect due to clustering

B. Query runs slower because clustering adds overhead

C. Query runs faster because data is clustered by region

D. Query performance is unchanged

Solution

1. Step 1: Understand clustering effect on filtering

   Clustering by region groups data physically, so filtering by region scans fewer data blocks.

2. Step 2: Analyze query impact

   The query filters on region = 'West', so it benefits from clustering, making it faster.

3. Final Answer:

   Query runs faster because data is clustered by region -> Option C

4. Quick Check:

   Filter on clustering key = faster query [OK]

Hint: Filtering on clustering key speeds queries [OK]

Common Mistakes:

• Thinking clustering slows queries
• Assuming clustering changes query results
• Believing clustering has no effect on performance

4.

Identify the error in this Snowflake command to add a clustering key on customer_id to an existing table orders:

ALTER TABLE orders ADD CLUSTER BY (customer_id);

medium

A. Should use ALTER TABLE orders CLUSTER BY (customer_id);

B. Cannot add clustering keys to existing tables

C. Clustering keys must be added during table creation only

D. Missing parentheses around customer_id

Solution

1. Step 1: Check Snowflake syntax for altering clustering keys

   Snowflake requires ALTER TABLE ... CLUSTER BY (columns) to add or change clustering keys.

2. Step 2: Compare given command

   The command uses ADD CLUSTER BY which is invalid syntax.

3. Final Answer:

   Should use ALTER TABLE orders CLUSTER BY (customer_id); -> Option A

4. Quick Check:

   Use CLUSTER BY with parentheses to alter keys [OK]

Hint: Use 'ALTER TABLE ... CLUSTER BY (cols)' to add keys [OK]

Common Mistakes:

• Using ADD instead of CLUSTER BY
• Omitting parentheses around columns
• Believing clustering keys can't be altered

5.

You have a large table events with columns event_type, event_date, and user_id. You often run queries filtering by event_type and event_date. Which clustering key choice will best improve query speed?

hard

A. CLUSTER BY (user_id)

B. CLUSTER BY (event_type, event_date)

C. CLUSTER BY (event_date)

D. No clustering key needed

Solution

1. Step 1: Identify columns used in frequent filters

   Queries filter on both event_type and event_date, so clustering on both helps.

2. Step 2: Evaluate clustering options

   CLUSTER BY (event_type, event_date) clusters by both columns, improving filtering speed. CLUSTER BY (user_id) clusters on unrelated user_id. CLUSTER BY (event_date) clusters only on one filter column. No clustering key needed ignores clustering benefits.

3. Final Answer:

   CLUSTER BY (event_type, event_date) -> Option B

4. Quick Check:

   Cluster on all filter columns for best speed [OK]

Hint: Cluster on all frequent filter columns together [OK]

Common Mistakes:

• Clustering on unrelated columns
• Clustering on only one of multiple filter columns
• Skipping clustering for large tables