Question #554
A company operates an on-premises data processing system configured with full redundancy across 15 servers in their data center. The system executes scheduled jobs that run every 4 hours and daily, in addition to ad-hoc user requests. Scheduled jobs require between 30 minutes and 3 hours to complete and must adhere to strict SLAs, accounting for 70% of system usage. User jobs typically finish within 10 minutes and can tolerate delays during system issues, making up the remaining 30% of usage. During outages, scheduled jobs must still meet their SLAs, while user jobs can be postponed.
The company aims to migrate this system to Amazon EC2 instances, adopting a consumption-based pricing model to minimize costs without long-term commitments. The solution must ensure high availability and uphold the SLAs for scheduled jobs.
Which solution MOST cost-effectively meets these requirements?
Distribute the 15 instances across three Availability Zones. In each zone, run three On-Demand Instances with Capacity Reservations and two Spot Instances.
Distribute the 15 instances across two Availability Zones. Run five On-Demand Instances with Capacity Reservations in each zone, and five Spot Instances in a third zone.
Distribute the 15 instances across three Availability Zones. In each zone, run four On-Demand Instances with Capacity Reservations and one Spot Instance.
Distribute the 15 instances across three Availability Zones. Use Savings Plans for six On-Demand Instances spread equally across zones, and deploy the remaining nine as Spot Instances.
Explanation
Option C is correct because:
1. High Availability (HA): Distributing instances across three AZs ensures redundancy. If one AZ fails, the remaining two AZs retain 8 On-Demand instances, maintaining SLA compliance for scheduled jobs.
2. Capacity Reservations: 4 On-Demand Instances per AZ with Capacity Reservations guarantee capacity for critical scheduled jobs, meeting strict SLAs even during outages.
3. Cost Optimization: Spot Instances (1 per AZ) handle user jobs, which can tolerate delays, reducing costs without long-term commitments.
4. No Savings Plans: Option D uses Savings Plans, which require commitments, conflicting with the requirement to avoid long-term contracts.
Other options fail because:
- A: 3 On-Demand per AZ may not provide sufficient reserved capacity during outages.
- B: Two AZs risk insufficient redundancy; Spot in a third zone may not align with HA requirements.
- D: Savings Plans involve commitments, and 6 On-Demand instances may not meet SLA needs during outages.
Key Points:
- Use Capacity Reservations for SLA-bound workloads.
- Spot Instances for interruptible, cost-sensitive tasks.
- Distribute across three AZs for HA.
Answer
The correct answer is: C