Improving Price Performance by 15%–20% Using AWS Graviton with DoubleCloud

DoubleCloud offers platform-as-a-service open-source technologies to help organizations streamline their data analytics infrastructure. It supplies fully managed solutions at scale to customers across the marketing, gaming, IT, and security industries. With purpose-built solutions, DoubleCloud helps customers run large database workloads, often ranging from hundreds of gigabytes to hundreds of terabytes in volume, without additional infrastructure setup.

In 2022, the company launched its managed services for ClickHouse, an online analytical processing database trusted by thousands of companies for near-real-time analytics, and for Apache Kafka, a data streaming technology for large-scale, data-intensive applications. DoubleCloud also launched a business intelligence service for dashboard analytics and an extract, transform, load service for data transfer. Using DoubleCloud’s Managed Service for ClickHouse and Managed Service for Apache Kafka, customers can customize and configure different compute resources based on their workloads. DoubleCloud’s fully managed solutions come with built-in integrations, auto scaling, and an engineer-friendly environment to make database deployment seamless.

Initially, DoubleCloud offered x86-based instances with its managed services, but the company sought other options to provide more flexibility for its customers. To focus on its value proposition of price performance, DoubleCloud decided to adopt AWS Graviton–based instances for its data plane to make its solution more performant for customers. It also decreased company costs by applying AWS Graviton–based instances to its control plane for service provisioning and management.

“We chose AWS Graviton–based instances to decrease compute costs and increase price performance efficiency,” says Victor Kiselev, head of product at DoubleCloud. “After running internal tests on our workloads, it was an easy decision.

During an initial benchmark test, AWS Graviton–based instances performed either on par or faster than the comparable non-AWS Graviton–based instances while being 15–20 percent more cost effective. After the company saw these benchmarking results, it moved forward with implementation in 2023. The team at DoubleCloud worked quickly to rearchitect its backend, adjusting its various Amazon EC2 components for collecting metrics and logging. Then, it integrated AWS Graviton–based instances with ease.

“Adopting AWS Graviton–based instances was a straightforward and smooth process,” says Misha Epikhin, head of product development at DoubleCloud. “We spent just 2–3 months with our team preparing our data plane for new CPU architecture. During that interval, we repackaged and rebuilt all packages and internal systems, including the databases, and deployed observability stacks.”

Customers now have the option of running their workloads on AWS Graviton–based instances, along with a combination of other services. Since the launch of AWS Graviton–based managed services in 2023, the percentage of customers’ ClickHouse and Apache Kafka clusters deployed using AWS Graviton has reached 45 percent. Having agility is crucial for DoubleCloud’s customers, who transfer 3–5 MB per second on average from different sources using Managed Service for Apache Kafka.

“When we onboard our customers, we pay close attention to latency metrics, how fast the ingestion process is, how fast queries are running, and what combination of processors they’re using,” says Kiselev. “Then, we take that information and help them fine-tune the solution according to their workloads.”

An additional customizable feature that DoubleCloud offers is the option to split data into various storage tiers to save customers up to six to eight times in costs. DoubleCloud powers this storage functionality using Amazon Simple Storage Service (Amazon S3), an object storage service offering scalability, data availability, security, and performance. Some customers opt to store more historical data in cold data storage classes offered through Amazon S3, which lowers overall storage costs. The data allocation is all integrated under the hood on ClickHouse so that customers do not need to worry about configuration.

“By adopting AWS Graviton–based instances, we’ve amplified efficiencies for our customers,” says Kiselev. “We can deliver optimal solutions for complex issues.”

AWS Graviton–based instances have become the first choice among DoubleCloud’s customers. The company continues to work on more advanced benchmarking under different scenarios and configurations for its AWS Graviton–based instances. It aims to keep improving its ClickHouse and Apache Kafka solutions to deliver exceptional performance, cost efficiency, and improved analytics for its customers.

“Customers are more receptive to implementing new solutions like AWS Graviton processors because we fully manage everything,” says Kiselev. “Our customers’ trust in our managed solutions lowers the barrier to adopting innovative new technologies.”