Electrolux Group · Appliances & High-Throughput Systems
Scaling thousands of operations with Kotlin Ktor microservices on AWS, fast, available, and resilient.
The Client
Electrolux is a global appliance company founded in 1919 in Stockholm. They sell in over 120 markets under brands like Electrolux, AEG, and Frigidaire, with around 50,000 employees worldwide.
Founded
1919
Headquarters
Stockholm
Project Overview
We built a high-throughput backend on Kotlin Ktor and AWS, with Kafka handling the event flow between services. The focus was performance at every level.
The Challenge
Re-engineer the stack for raw throughput.
The existing platform ran on Kotlin and Java Spring Boot, with slow startup times, heavy memory usage, and too much boilerplate. Every optimization cycle was expensive. Electrolux needed to handle thousands of concurrent operations across global markets with sub-second response times, but the stack wasn't built for that kind of throughput. Performance had to improve across the board: cold starts, serialization, connection pooling, everything.
Thousands of concurrent operations with sub-second P99 response times.
The Solution
Lean, coroutine-native microservices.
We replaced Spring Boot with Kotlin Ktor, which is lightweight, coroutine-native, and built for concurrency. Kafka decouples the services and handles event delivery. Kubernetes on EKS auto-scales the fleet, and Terraform keeps infrastructure reproducible. Circuit breakers and bulkheads keep things running when parts fail.
Technology Stack
Ktor microservices on AWS, Kafka for events, Kubernetes for orchestration, and Terraform for reproducible infrastructure.
Kotlin Ktor
We picked Ktor because it's lean. No annotation magic and no heavyweight container, just coroutines and explicit routing. Each service handles thousands of connections with minimal threads. Startup is fast, memory is low, and the code stays readable.
Results & Impact
Built for throughput.
- Thousands of concurrent operations with sub-second P99 response times
- 99.99% uptime with multi-AZ, circuit breakers, and self-healing pods
- Millions of Kafka events processed daily with exactly-once delivery
- 40% lower compute costs through auto-scaling during off-peak hours
- Zero-downtime deployments via rolling updates
- 10x more concurrent connections per instance compared to the old Spring Boot stack
10x
More concurrent connections
99.99%
Platform uptime
40%
Lower compute cost
Ready to scale your operations?
Need a backend that actually scales? Let's talk about the performance you need.