Drift Case study

Drift Reduces EKS Costs by 15% with Intel Granulate Capacity Optimization

Download Case Study

The Role of Compute at Drift

Infrastructure: Java on Amazon EKS

Drift’s engineering teams tackle some of the most meaningful problems in the industry—from building adaptive AI for bot-supported conversations, to designing next generation chat experiences, to bridging the gap between old and new ways to buy.

The team is responsible for guiding architectural decisions and owns the tech stack within their respective products. As the organization continues to grow, the department is looking for their systems and applications to scale and support their rapidly growing customer base.

Developers and engineers at Drift create infrastructure that processes billions of messages a minute and extract buyer intent and priority for conversations. They also build tools to trace and visualize any message ever sent to their customers so that they can zero in on giving the best buying experience.  

Drift’s Kubernetes Optimization Journey

At Drift, the process of enhancing and optimizing their EKS infrastructure was heavily reliant on manual efforts and hands-on management. Given the constraints, especially with their limited engineering bandwidth, the team at Drift was actively in search of an approach that would yield significant cost savings without requiring an intensive input of resources or time. 

To address this challenge, they executed a strategic deployment of both spot and reserved instances. This combination was not just about diversifying their instance usage but also about proactively mitigating the often burdensome on-demand charges. Such a move was indicative of Drift's commitment to operational efficiency while maintaining fiscal responsibility.

Ultimately, despite their prior initiatives, Drift continued to actively seek out approaches that were both high impact and required minimal effort, all in a bid to achieve more substantial cost savings in their operations.

We have already seen truly impressive results from Granulate on Kubernetes cost and performance with no engineering requirements. We see more potential for cost reduction and are looking forward to implementing Granulate for further savings.
Matt Jackson, DevOps Director

Results of Intel Granulate’s Capacity Optimization 

Drift began by deploying Granulate’s Continuous Profiler to establish benchmarks and discover code bottlenecks. After a two week benchmarking period, the Granulate agent identified compelling optimization opportunities on both the Java and Kubernetes layers.

Upon employing Granulate’s Capacity Optimization solutions, one primary production cluster comprising 4,000 cores was autonomously optimized. The immediate outcome was an impressive 15% cost reduction in Drift’s main production environment.

Capacity Optimization consisted of autonomous orchestration for pods CPU, memory and HPA definitions with continuous auto-pilot. The solution eliminates over-provisioning of EKS clusters, thereby reducing infrastructure costs with guaranteed SLA.

20.24% Improvement in Reserved Cores

15.46% Improvement in Node Count

While the full potential of cost reduction is yet to be realized, a promising avenue remains. Pending reliability testing, Intel Granulate’s Capacity Optimization autopilot solution projects potential of up to 48% of savings.

*Intel does not control or audit third-party data. You should consult other sources to evaluate accuracy


HQ: Boston, USA
Employees: 612

Drift is a Conversational Marketing platform that combines chat, email, video, and AI to remove the friction from business buying. With Drift, you can start conversations with future customers now, on their terms — not days later.

There are over 50,000 businesses that use Drift today to accelerate revenue, shrink sales cycles, and make buying easy. Drift’s mission is to use conversations to make business buying frictionless, more enjoyable, and more human.

Get an evaluation of your workload performance improvements

Talk to expert