The Client
McQueen Autocorp, a US-based used car selling platform, connects sellers with a nationwide network of buyers. Utilizing price prediction algorithms and real-time bidding, it provides instant offers and manages the entire sales process.
Disclaimer: The name McQueen Autocorp is a placeholder, as both parties have signed an NDA.
The Challenge
As McQueen Autocorp scaled, Kubernetes workloads on x86_64 architecture drove up costs. Key challenges included:
- High compute costs.
- Risk of downtime during migration.
- Applications built only for x86_64.
- CI/CD pipelines lacked multi-platform builds.
- Balancing cost savings with performance.
The Solution
We executed a phased migration strategy focused on efficiency:
- Migrated workloads to Graviton-enabled Kubernetes nodes.
- Built multi-platform container images.
- Enhanced CI/CD pipelines for multi-architecture support.
- Allowed developers to toggle Graviton.
- Preserved performance with zero disruption.
Results
Our collaboration delivered measurable outcomes:
- A 15% reduction in daily EC2 costs post-migration.
- Seamless migration with no downtime.
- Flexible architecture for x86_64 and Graviton.
- Faster deployments via updated CI/CD pipelines.
- Improved efficiency with plans for full Graviton adoption.
