When one of our enterprise clients asked us to build a contact center platform capable of handling 50 million interactions per month, we knew we needed an architecture designed for failure, not one that assumed everything would go right.
Starting With the Right Foundation
We chose AWS as our primary cloud provider for its breadth of managed services and global reach. The core stack: EKS for container orchestration, Aurora PostgreSQL for the primary database, ElastiCache Redis for session state, and SQS for decoupled message processing.
Everything was defined in Terraform from day one. Infrastructure as Code was non-negotiable. It gave us reproducible environments, meaningful code reviews for infrastructure changes, and the ability to spin up isolated staging environments on demand.
The Kubernetes Layer
Our Kubernetes setup across three availability zones gave us the fault tolerance the SLA demanded. Each service was independently scalable: the transcription service, the routing engine, and the analytics pipeline could all scale horizontally without touching each other.
We implemented Horizontal Pod Autoscaling based on custom metrics from CloudWatch, which meant the cluster could respond to traffic spikes within 90 seconds. This was critical during peak periods when interaction volume could double within minutes.
Achieving 99.9% Uptime
The 99.9% uptime SLA translates to less than 9 hours of downtime per year. To achieve this, we implemented circuit breakers on all service-to-service calls, blue/green deployments for zero-downtime releases, and multi-region failover for the most critical components.
Observability was built in from the start, distributed tracing with AWS X-Ray, structured logging in CloudWatch, and custom dashboards in Grafana that gave the on-call team everything they needed to diagnose and resolve issues in minutes, not hours.
Put this to work in your organisation
Talk to our experts and get a free consultation tailored to your needs.
