Skip to main content

ClickHouse delivers faster analytics at scale using AWS Graviton instances

Learn how real-time database company ClickHouse improved database performance and reduced costs by adopting AWS Graviton–based instances.

Key Outcomes

64%
more queries per second using Graviton4 instances compared to x86-based instances
25%
query performance improvement on Graviton3 instances compared to x86-based instances
36%
performance boost in benchmarks for Graviton5 instances compared to Graviton4 instances
16%
higher concurrency for Graviton5 instances compared to Graviton4 instances

Overview

As its analytic workloads grew, ClickHouse wanted to improve the price performance of its service—ClickHouse Cloud—without adding operational complexity. At the same time, new usage patterns such as AI-driven workflows were increasing query concurrency demands. To deliver fast, interactive analytics at scale while maintaining cost predictability, ClickHouse worked alongside Amazon Web Services (AWS) to enhance efficiency. By optimizing the compute powering its analytics workloads, the company was able to reinvest gains to deliver even faster analytics experiences for customers.

About ClickHouse

ClickHouse is a leading high-performance data platform, powering workloads that span near real-time analytics, data warehousing, observability, and AI/ML.

Opportunity | Using Graviton-based instances to scale analytics

ClickHouse, now an AWS Partner, began as a self-managed, open source database for online analytics processing. As adoption grew, the team formed a company around the project and later introduced ClickHouse Cloud, a fully managed service available through AWS Marketplace. This solution runs workloads with enhanced performance and developer experience in the cloud.

ClickHouse Cloud initially ran on general-purpose and memory-optimized x86-based Amazon Elastic Compute Cloud (Amazon EC2) instances, which provided secure and resizable compute capacity. But as AI-driven workflows increased query density, ClickHouse needed to balance compute capacity with cost efficiency while meeting performance expectations. “AI agents query databases to provide responses,” says Alexey Milovidov, founder and chief technology officer of ClickHouse. “If every query takes 10 seconds, then dozens of queries take minutes, and this leads to untenable latency.”

To improve compute efficiency, ClickHouse began evaluating AWS Graviton–based Amazon EC2 Instances. These are custom-designed server processors that were developed by AWS to provide up to 40 percent better price performance for cloud workloads running on Amazon EC2.

Solution | Running ClickHouse Cloud on Graviton–based instances

To orchestrate compute across regions and instance types, the company operates ClickHouse Cloud on Amazon Elastic Kubernetes Service (Amazon EKS), a service for starting, running, and scaling Kubernetes. ClickHouse stores data in Amazon Simple Storage Service (Amazon S3), an object storage service built to store and retrieve virtually any amount of data from anywhere. By separating compute from storage, the company can change instance types without migrating data or redesigning the architecture.

After benchmarking Graviton-based instances against existing x86 instances, ClickHouse migrated production workloads to Graviton3-based instances. The initial transition required careful planning and testing, including upgrading to newer virtual machine generations. But after establishing Graviton3-based instances in production, ClickHouse migrated to Graviton4-based instances with minimal effort. “The complexity was almost zero,” says Milovidov. “We didn’t make any changes in the code base for switching.”

Because ClickHouse is an AWS Partner, the company worked alongside AWS to benchmark new generations and even plan capacity across regions. “I like the quality of AWS support,” says Milovidov. “We get information about upcoming releases so that we can begin early testing and preparation, and we often consult about the availability in different regions.” This collaboration helped ClickHouse align its rollout plans with new AWS compute offerings as they launched.

“Graviton-based instances became the default compute option for ClickHouse Cloud,” says Timur Solodovnikov, principal software engineer at ClickHouse. This way, ClickHouse has a repeatable process for adopting newer generations as they become available. “We could not have achieved that level of efficiency without the help of the AWS team,” says Krithika Balagurunathan, senior director of product management at ClickHouse.

Outcome | Boosting efficiency and price performance at scale

Compared with similar x86-based instances, Graviton3-based instances improved query performance by 25 percent, while Graviton4-based instances increased queries per second by 64 percent for certain workloads. What’s more, early benchmarking of Graviton5-based instances showed a 36 percent performance boost, with 16 percent higher concurrency compared to the previous generation. Today, the company can maintain cost stability as query volumes grow and has achieved a 2 percent cost reduction per unit for ClickHouse credits. “The migration to Graviton-based instances delivered cost savings, and we use those savings to provide better service for our customers,” says Milovidov.

ClickHouse’s architecture on AWS supports both ClickHouse Cloud and self-managed deployments, and the same performance benefits apply to customers running ClickHouse on Graviton-based instances. As workloads evolve, ClickHouse will continue to work alongside AWS and adopt new Graviton-based instance generations, strengthening a long-term cloud strategy to improve efficiency. 

Missing alt text value
The migration to Graviton-based instances delivered cost savings, and we use those savings to provide better service for our customers.

Alexey Milovidov

Founder and Chief Technology Officer, ClickHouse

Did you find what you were looking for today?

Let us know so we can improve the quality of the content on our pages