An on-premises analytics platform runs across 12 servers in a fully redundant configuration. It handles scheduled jobs (hourly and daily) that take 20 minutes to 2 hours and have strict SLAs (65% of load) and ad-hoc user jobs that finish within 5 minutes and have no SLA (35% of load). During failures, scheduled jobs must preserve SLAs while user jobs can be delayed. You must move to EC2 and use consumption-based pricing with no long-term commitments, maintaining high availability and SLAs for scheduled jobs at minimal cost. Which option is the most cost-effective while meeting requirements?

The correct option ensures high availability and SLA for scheduled jobs at minimal cost. Distributing 12 instances across three Availability Zones (AZs) with three On-Demand Instances with Capacity Reservations per AZ (total 9) provides redundancy and guaranteed capacity for the critical 65% scheduled workload. On-Demand Instances with Capacity Reservations are ideal for strict SLAs as they guarantee capacity and are not interrupted. The remaining one Spot Instance per AZ (total 3) handles the non-SLA ad-hoc jobs, leveraging cost savings from Spot Instances which can be interrupted. Incorrect options: Running two On-Demand with Capacity Reservations and four Spot Instances per AZ (total 4 On-Demand, 8 Spot) might not guarantee enough capacity for scheduled jobs if Spot Instances are reclaimed, and two AZs offer less redundancy than three. Running all four instances in one AZ as On-Demand with Capacity Reservations and the rest as Spot Instances concentrates critical capacity in a single AZ, reducing overall availability. Using Savings Plans with On-Demand Instances provides cost savings but doesn't guarantee capacity like Capacity Reservations, which is crucial for strict SLAs.