AWS Certified Solutions Architect - Professional

Implement AWS App Auto Scaling for ECS instead of using Kubernetes native autoscaling, migrate services to ECS Fargate for better scaling performance, and use Target Tracking Scaling policies based on CloudWatch metrics.

Use Kubernetes Cluster Autoscaler with default CPU-based Horizontal Pod Autoscaler for all services. Set aggressive scaling policies (scale up at 50% CPU, scale down at 20% CPU) to respond quickly to traffic spikes.

Deploy Karpenter (AWS-native Kubernetes autoscaler) to replace Cluster Autoscaler for faster node provisioning, use KEDA (Kubernetes Event-Driven Autoscaling) with scalers for different metrics (HTTP requests, SQS queue, memory), and implement pod priority classes to ensure critical services get resources first during scale events.

Implement Horizontal Pod Autoscaler (HPA) with custom metrics for each service: HPA based on ALB requests per second for frontend services (using AWS Load Balancer Controller metrics), HPA based on SQS queue depth for order processing (using CloudWatch Container Insights), and HPA based on memory utilization for recommendation service. Configure Cluster Autoscaler for node scaling and implement PodDisruptionBudgets to prevent cascade failures. Pre-scale infrastructure 2 weeks before Black Friday using scheduled scaling.