How to Get Started with ClickHouse Cloud

ClickHouse Cloud lets analysts and operators run sub-second SQL queries on large datasets without managing servers. The core idea: instead of rows, ClickHouse stores data column by column, so a query that scans one column across 500 million rows only reads that column rather than every field in every row. That design difference is why ClickHouse benchmarks show 10 to 100 times faster query execution than row-oriented databases like PostgreSQL for aggregation workloads.

This guide covers how to get a ClickHouse Cloud service running, load your first dataset, write queries, and connect it to a BI tool.

Why ClickHouse Instead of Your Existing Database

Standard databases such as PostgreSQL or MySQL are row-oriented, meaning all columns for a row are stored together on disk. That layout is efficient for transactional operations that read one record at a time, but slow for analytics that aggregate across millions of records.

ClickHouse stores each column separately. A query like SELECT region, SUM(revenue) FROM orders GROUP BY region only needs to read two columns out of however many your orders table has. On a table with 20 columns and 100 million rows, that is an order-of-magnitude reduction in data read from disk.

Practical benchmark numbers from March 2026: ClickHouse Cloud ranked first on ClickBench across all systems tested, with 30 to 55 percent faster query performance than competing managed services. In star schema join tests across six tables at scales between 721 million and 7.2 billion rows, ClickHouse Cloud was faster and cheaper than Snowflake and Databricks at every scale with no manual tuning.

The trade-off: ClickHouse is optimized for reads and bulk inserts, not for frequent single-row updates or transactions. It is the right tool for reporting, dashboards, log analysis, and event data. It is the wrong tool for an order management system that updates individual records constantly.

Creating a ClickHouse Cloud Account

Go to clickhouse.com and sign up for a free trial. The free tier gives you a Development service with enough capacity to load and query datasets up to a few hundred gigabytes.

After sign-up, the console asks you to create your first service. Choose a cloud provider and region closest to where your data currently lives. Give the service a name and click Create. The service takes about 90 seconds to initialize.

When the service is ready, you will see a connection string in the console. Save the host URL, username, and password. You will need these to connect external tools later.

Loading Your First Data

The simplest path is to upload a CSV file directly through the ClickHouse Cloud console.

Open the SQL console from the left sidebar. To create a table from a CSV, click the Import icon and select Upload a file. ClickHouse will infer column types from the first 100 rows and propose a CREATE TABLE statement. Review the types, adjust any columns that were inferred incorrectly (dates often need manual correction), and click Create Table and Import Data.

For programmatic loads, ClickHouse exposes an HTTP interface. The following command inserts a CSV file directly:

curl -X POST "https://YOUR_HOST:8443/" \
  --user "default:YOUR_PASSWORD" \
  --data-binary @orders.csv \
  "?query=INSERT INTO orders FORMAT CSVWithNames"

For ongoing ingestion from a data warehouse or SaaS source, ClickHouse has native connectors for S3, PostgreSQL, MySQL, Kafka, and most major platforms. The S3 connector is the most common path for teams that already land data in object storage:

INSERT INTO orders
SELECT * FROM s3(
  's3://your-bucket/orders/*.parquet',
  'AWS_KEY', 'AWS_SECRET',
  'Parquet'
);

Running Your First Analytics Queries

Once data is loaded, open the SQL console and try a basic aggregation:

SELECT
  toStartOfMonth(order_date) AS month,
  region,
  COUNT()                    AS orders,
  SUM(revenue)               AS total_revenue
FROM orders
WHERE order_date >= '2025-01-01'
GROUP BY month, region
ORDER BY month DESC, total_revenue DESC
LIMIT 50;

On a table with 50 million rows, this query typically returns in under two seconds on a Development service. On a Production service, it returns in milliseconds.

ClickHouse SQL is mostly standard with a few additions worth knowing. The toStartOfMonth(), toStartOfWeek(), and toStartOfDay() functions make time-series aggregations concise. The countIf() and sumIf() functions let you add conditional aggregations without subqueries:

SELECT
  region,
  countIf(status = 'completed')  AS completed_orders,
  sumIf(revenue, status = 'completed') AS completed_revenue
FROM orders
GROUP BY region;

The arrayJoin() function handles JSON arrays and nested fields that would require a lateral join in PostgreSQL.

Connecting to a BI Tool

ClickHouse Cloud works with most modern BI tools via a native ClickHouse connector or via the MySQL-compatible interface.

For Grafana, install the official ClickHouse data source plugin, enter your host URL and credentials, and write queries directly in the query editor. Grafana's time series panels work well with ClickHouse's time functions.

For Tableau, use the ClickHouse ODBC driver. Download it from the ClickHouse integrations page, configure a DSN with your host and credentials, and connect Tableau using Other Databases (ODBC).

For Metabase, ClickHouse is available as a native database type in recent versions. Add it under Admin, Databases, and supply the host, port (8123 for HTTP, 9440 for HTTPS native), username, and password.

For tools that do not have a native connector, ClickHouse exposes a MySQL-compatible interface on port 9004. Any tool that can connect to MySQL can connect to ClickHouse using the same credentials. Enable the MySQL interface in your service settings under Network, MySQL Interface.

If you want to skip SQL entirely and ask questions about your data in plain English, VSLZ handles that from a file upload with no ClickHouse configuration needed.

Managing Costs on ClickHouse Cloud

ClickHouse Cloud bills on compute time plus storage. The Development tier is the most economical for exploration because it scales to zero when idle. Production tiers have minimum replica counts that prevent scale-to-zero but guarantee low latency.

The two biggest cost drivers to watch: full table scans on large tables and high-cardinality ORDER BY clauses. Use the EXPLAIN command to see what ClickHouse reads before running a heavy query:

EXPLAIN SELECT region, SUM(revenue) FROM orders GROUP BY region;

Look for Granules in the output. Each granule is an 8,192-row block. Fewer granules read means lower cost.

Primary keys in ClickHouse are sparse indexes that determine physical sort order on disk. Choosing a primary key that matches your most common filter column (for example, ORDER BY (region, order_date) if most queries filter by region) dramatically reduces the granules scanned per query.

Next Steps

After loading your first dataset and validating query performance, the natural next move is to set up an incremental ingestion pipeline so ClickHouse stays current with your operational data. Most teams use Airbyte, Fivetran, or dbt to manage that layer. ClickHouse's ReplacingMergeTree engine handles deduplication for sources that emit updates, so incremental loads do not require DELETE operations.

ClickHouse Cloud's query API lets you expose any SQL query as a REST endpoint, which is useful if you want to feed dashboard metrics directly from ClickHouse into a frontend or Slack bot without a separate API layer.