Rye: 12% savings and more reliable operation through AWS migration

"At Rye, our production infrastructure ran primarily on GCP — Cloud Run, Cloud Tasks, Pub/Sub, Firestore, and GCE — but our engineering team's deeper expertise is in AWS. We partnered with Tech 42 to migrate our platform to AWS.

Tech 42 designed and built a full multi-account AWS foundation for us: AWS Organizations with Control Tower governance, isolated Dev, Staging, and Production environments, and networking spanning multiple Availability Zones. They containerized and migrated our application services to ECS on Fargate, and moved our asynchronous workloads to Lambda functions integrated directly with SQS, with EventBridge-driven scheduling for our recurring jobs. They also connected our AWS environment to MongoDB Atlas over VPC peering, stood up CI/CD pipelines for both application deployments and Terraform-managed infrastructure, and built out a full observability stack shipping logs, metrics, and traces to Grafana Cloud. Everything was delivered as Terraform in our own repositories, along with a full README and architecture diagram, so our team can operate and extend the environment ourselves going forward.

Migrating from Cloud Run to AWS ECS will save us roughly 12% on our compute costs by letting us binpack services more efficiently. Cold start times for our Node.js services on Cloud Run were extremely long — 40+ seconds in the worst case — which prevented us from elastically scaling our services; we're seeing a 75%+ improvement in cold starts on AWS, which will let us cut our compute footprint another ~20% through right-sizing our instance counts. We'd also had two production incidents last year caused by contention on Cloud Tasks queues, since they don't natively support fair queuing — we're excited to completely eliminate this failure mode going forward with SQS's fair queuing capabilities.

Tech 42 also walked us through a recorded knowledge transfer session, which gave our team a clear, hands-on understanding of how to deploy and manage the environment going forward. Communication throughout the engagement was consistent and responsive — the team was quick to help us work through issues on Slack, proactive about keeping testing on track, and flexible whenever we needed to adjust scope. We're happy with how the engagement went and are looking forward to continuing to work with Tech 42 as we plan our next phase on AWS."