An application runs on an Auto Scaling group of On-Demand EC2 instances behind an ALB. The workload contains critical analysis that cannot tolerate interruptions and must scale quickly, and noncritical analysis that can tolerate interruptions and is memory-intensive. You must reduce scale-out latency for the critical workload while also processing the noncritical workload. Which combination of steps meets these requirements? (Choose two.)

The first correct option addresses the critical workload. A warm pool of stopped instances significantly reduces scale-out latency because instances are pre-provisioned and only need to start up, rather than launch from scratch. Using On-Demand Instances ensures no interruptions, as required for critical analysis. Detailed monitoring is a good practice but not the primary driver for latency reduction here. The first incorrect option suggests Spot Instances, which are interruptible and unsuitable for critical workloads. The second correct option addresses the noncritical workload. Creating a separate Auto Scaling group allows independent scaling. Using Spot Instances is cost-effective for noncritical, interruptible workloads. Since the noncritical workload is memory-intensive, configuring a custom memory utilization metric via the CloudWatch agent is essential for accurate scaling, as EC2 does not provide memory metrics by default. Modifying the application to use two target groups allows the ALB to route critical and noncritical traffic to their respective Auto Scaling groups. The second incorrect option is wrong because there is no predefined memory utilization metric type for target-tracking scaling policies without the CloudWatch agent.