A company runs a custom application on Amazon EC2 On-Demand Instances. The application has frontend nodes that must run 24/7. The backend nodes only need to run for short periods depending on the workload. Frontend nodes accept jobs and place them in queues. Backend nodes asynchronously process jobs from the queues, and jobs can be restarted. The company wants to scale infrastructure based on workload, using the most cost-effective option. Which solution meets these requirements MOST cost-effectively?