Explanation
Correct Answer: D - Use AWS Batch on Amazon EC2
AWS Batch is specifically designed for running batch computing workloads on AWS. Here's why this is the best choice:
Key Factors:
- CPU-intensive workload: The application requires 64 vCPU and 512 GiB memory, which is a substantial compute requirement.
- Batch job nature: Runs every hour for 15 minutes - this is a classic batch processing pattern.
- Least operational overhead: AWS Batch manages the provisioning, scheduling, and scaling of compute resources.
Why AWS Batch is optimal:
- Automatic resource provisioning: AWS Batch automatically provisions EC2 instances with the required specifications (64 vCPU, 512 GiB memory)
- Job scheduling: Handles job queuing and scheduling automatically
- Cost optimization: Only runs instances when jobs need to execute, then terminates them
- Minimal management: No need to manage clusters, scaling policies, or instance lifecycle
- Integration: Works seamlessly with other AWS services
Why other options are less optimal:
A. AWS Lambda:
- Lambda has limitations (15-minute timeout, memory up to 10GB, CPU scaling based on memory)
- Not suitable for 64 vCPU workloads
- Would require complex parallelization and coordination
B. Amazon ECS with Fargate:
- Fargate has resource limits (currently up to 16 vCPU and 120GB memory per task)
- Would require multiple tasks to achieve 64 vCPU
- More operational overhead for scheduling and coordination
C. Amazon Lightsail with Auto Scaling:
- Lightsail is designed for simpler workloads
- Auto Scaling adds operational complexity
- Not optimized for batch processing
D. AWS Batch on EC2:
- ✓ Specifically designed for batch workloads
- ✓ Can provision large EC2 instances (up to 448 vCPU, 24TB memory)
- ✓ Minimal operational overhead - AWS manages scheduling and provisioning
- ✓ Cost-effective - only pay for compute when jobs run
Additional Considerations:
- AWS Batch can use Spot Instances for additional cost savings
- Job dependencies and retries are built-in
- Monitoring and logging integration with CloudWatch
- Supports Docker containers for application packaging
This solution provides the required compute power while minimizing operational overhead, making it the most appropriate choice for this CPU-intensive batch workload.
