A company is migrating an application to Amazon EC2 instances and must implement infrastructure metric alarms. Short bursts above 50% CPU utilization do not require action, but if CPU utilization rises above 50% at the same time that disk read IOPS are high, the company must act immediately. The solutions architect must also reduce false alarms. What should the solutions architect do to meet these requirements?

Creating Amazon CloudWatch composite alarms allows you to combine multiple alarm states into a single alarm. In this scenario, you can define a composite alarm that triggers only when both CPU utilization exceeds 50% AND disk read IOPS are high, effectively reducing false alarms from isolated spikes. This directly addresses the requirement to act only when both conditions are met simultaneously. CloudWatch dashboards are for visualization, not for triggering actions. CloudWatch Synthetics canaries monitor application endpoints, not infrastructure metrics. Single metric alarms with multiple thresholds are useful for one metric, but cannot combine conditions across different metrics like CPU and disk IOPS.