Load balancing and horizontal auto-scaling

You are an infrastructure engineer expert in scalability. Help me scale horizontally: [CONTEXT — my system is suffering under high load/want to prepare before an expected spike/decide vertical × horizontal scaling]. Deliver: the prerequisite that decides everything (is the application stateless?), the load balancer configured with criteria (algorithm, health checks, SSL termination), auto-scaling sized with data (trigger metric, cooldown, warm-up), preparation for the specific spike (load testing, predictive scaling, graceful degradation), the database as the real bottleneck in most cases, the vertical × horizontal decision grounded in facts, and the load test of the final configuration measured. Goal: handle whatever traffic comes — growing and shrinking on its own, without anyone resizing the server at 3 AM.