Follow us
Breaking
Tech Services

Biggest trends in ClickHouse for analytics teams

ClickHouse is expanding its managed cloud services and Postgres integration to compete with Snowflake and BigQuery. Benchmarks show ClickHouse achieving a fresh-data-path cost of $3.49, significantly lower than Snowflake's $9.65 and BigQuery's $29.55.

Share

ClickHouse managed services and the cloud shift

ClickHouse Cloud provides a fully managed, SOC 2 Type 2-compliant version of the database on the AWS Marketplace. This service enables users to scale processing power without touching data storage through the separation of compute and storage. The company provides a lower pricing tier for developers after a 30-day free trial period. This version is available for Google Cloud and Microsoft Azure in the coming months. You already know that managing complex database systems requires significant engineering resources. Avery Kiemmer of Thrive Capital says ClickHouse is the most efficient database for fast and large-scale analytics. Aaron Katz, the CEO of ClickHouse, says the future of databases lies in managed services like serverless cloud offerings that are cloud agnostic and multi-cloud. Katz previously held roles at Salesforce and Elastic. The company began as an open-source project at Yandex to power Yandex.Metrica. After the spin-off in 2021, the company raised $50 million in Series A funding led by Index Ventures and Benchmark. ClickHouse also acquired Arctype to include a native GUI for managing and querying databases inside ClickHouse Cloud. The company is used by Cloudflare, Uber, eBay, Comcast, and Cisco.

The Postgres and ClickHouse integration trend

The acquisition of PeerDB enables ClickHouse to deliver specialized change data capture capabilities that allow companies to sync transactional data from Postgres into an analytical warehouse with significantly higher speeds. PeerDB co-founder Sai Srirampur, who worked on Azure’s PostgreSQL service, says the ClickHouse connector is the fastest-growing connector, surpassing Snowflake and BigQuery. PeerDB allows for the initial load of terabytes of data from a Postgres database to a warehouse. Many companies use Postgres as the transactional backend for applications and then move that data into ClickHouse for analytics. ClickHouse launched an enterprise-grade Postgres service in partnership with Ubicloud to combine transactional and analytical workloads on a single foundation. This integration allows developers to synchronize transactional data into ClickHouse with a few clicks and achieve analytics performance improvements of up to 100x. This service uses NVMe storage to support high-performance workloads. Ubicloud leaders include individuals from Citus Data, Heroku, and Microsoft. Will the combination of OLTP and OLAP on one foundation eliminate the need for seven different database types?

Comparing Snowflake, BigQuery, and ClickHouse

Analytics teams choose between Snowflake, BigQuery, and ClickHouse based on how they manage costs and concurrency. Enterprises store data across seven or more different databases, including Snowflake, ClickHouse, Timescale, Elastic, S3, Postgres, Redis, Cassandra, and Dgraph. Snowflake separates storage and compute using virtual warehouses to prevent resource conflicts. These virtual warehouses allow different teams to run queries without interference. Snowflake includes features like Time Travel, which allows users to view data as it existed up to 90 days in the past. Snowflake also provides secure data sharing to allow live, governed data sharing with partners. Snowflake includes Snowpark for building pipelines in Python, Java, Scala, and SQL. Snowflake costs increase if users leave active virtual warehouses running without pausing them. Snowflake compute costs depend on usage, with on-demand rates for credits and pre-purchased capacity for discounts. Snowflake storage costs $23 per TB per month on AWS US East at list price.

BigQuery uses a serverless framework built on Google technology like Dremel. It scales automatically and requires no manual infrastructure management. BigQuery ML allows users to build and run machine learning models using standard SQL. BigQuery also handles real-time ingestion through streaming inserts. BigQuery charges based on the amount of data scanned per query, usually $6.25 per terabyte. BigQuery costs increase when queries scan large datasets frequently. BigQuery also has a capacity option where users pay a flat rate for dedicated processing power, with the Standard Edition starting at $0.04 per slot hour. BigQuery storage costs $23 per TB per month for active storage, but the price drops to $16 per TB if a table remains unmodified for 90 days. The first 1 TiB scanned each month is free.

ClickHouse provides a different model where costs start at $50 per month and include a free option.

Feature ClickHouse Snowflake BigQuery
Pricing Model Usage-based Consumption-based Usage-based
Starting Cost $50/month Not listed First 1 TiB free
Free Tier Yes Not listed Yes
Infrastructure Self-managed or Cloud Fully managed Fully managed

Benchmarks show ClickHouse has a fresh-data-path cost of $3.49 and a query cost below $0.01 with a 5.6 second runtime. Snowflake has a fresh-data-path cost of $9.65 and a query cost of $0.76 with a 7.08 minute runtime. BigQuery has a fresh-data-path cost of $29.55 and a query cost of $0.55 with a 6.38 minute runtime.

Share

Technewsdaily

Senior tech writer covering AI, gadgets and cybersecurity. Breaking down the news that matters, every day.