A financial services company, faced rising costs from its serverless infrastructure.
The primary cost drivers were Aurora DB and AWS Lambda, necessitating drastic measures to achieve cost reductions well beyond the typical 15%-20% savings from standard optimizations.
The company required a strategic overhaul to slash costs by more than 50%, as they felt trapped without a clear path to achieving the necessary savings.
The challenge was to reduce operational costs significantly while maintaining scalability and performance during high traffic peaks.
Lambda Transformation:
We began by converting all Lambda functions into Docker containers, taking advantage of their self-contained nature to enhance manageability and cost-efficiency.
Hosting Strategy:
These containers were then hosted on EC2, using a mix of on-demand instances for baseline loads and spot instances for peak times, optimizing both cost and performance.
Aurora to Postgres Transition:
Switched from large Aurora instances to smaller Postgres servers with read replicas. We used a proxy for efficient read distribution. Then, we optimized cloud resource utilization with EC2 and Spot instances.
Replacing SNS with RabbitMQ:
To further reduce costs, we substituted AWS SNS with RabbitMQ, an open-source message queue system hosted on spot EC2 instances.
This switch decreased dependencies on higher-cost cloud services and enhanced overall system efficiency.
The strategic restructuring led to immediate and significant financial benefits:
After Phase 1:
We achieved a 50% reduction in costs from the initial serverless setup.
After Phase 2:
Further optimizations brought the costs down by an additional 20%, totaling a 70% reduction.
Final Adjustments:
The implementation of RabbitMQ contributed an additional 10% cost reduction, cumulatively saving over 80% annually.
