##### ###### ##### ### # # ### # # ###### ## ## ## ## ## ## ## # # # # # ## ##### #### ##### # # # # # # # #### ## # ## ## ## ## # # # # # ## ## # ###### ## ### # ### # ######
##### ###### ##### ### # # ### # # ###### ## ## ## ## ## ## ## # # # # # ## ##### #### ##### # # # # # # # #### ## # ## ## ## ## # # # # # ## ## # ###### ## ### # ### # ######
Auto-scaling adds or removes instances based on metrics (CPU, request count, queue depth). The load balancer must integrate seamlessly.
New instances register with the LB after passing health checks. Scaling down triggers connection draining before removal. Target tracking policies maintain a target metric (e.g., % CPU) automatically.
