Question #752
A company processes large-scale distributed data analysis jobs using AWS Batch. Each job is stateless, fault-tolerant, and runs for up to 4 hours.
Which pricing model should the company use to minimize costs while meeting these requirements?
Reserved Instances
Spot Instances
On-Demand Instances
Dedicated Instances
Explanation
The correct answer is B. Spot Instances.
Why Spot Instances?
- Cost Efficiency: Spot Instances offer up to 90% cost savings compared to On-Demand Instances, making them the cheapest option.
- Fault Tolerance: Since the jobs are stateless and fault-tolerant, Spot Instance interruptions (which can occur with short notice) can be handled gracefully by restarting the job on another instance.
- AWS Batch Integration: AWS Batch automatically retries failed jobs, making it well-suited for Spot Instances.
Why Other Options Are Incorrect:
- A. Reserved Instances (RIs): RIs require a long-term commitment (1-3 years) and are cost-effective only for steady-state workloads, not short-term, intermittent jobs.
- C. On-Demand Instances: These are more expensive than Spot Instances and unnecessary for fault-tolerant workloads.
- D. Dedicated Instances: These are physically isolated for compliance/security and provide no cost benefits over Spot Instances.
Key Points to Remember:
- Use Spot Instances for fault-tolerant, stateless, and short-lived workloads.
- AWS Batch handles Spot interruptions by resubmitting jobs.
- Avoid Reserved/Dedicated Instances unless required for specific compliance or long-term usage.
Answer
The correct answer is: B