The correct answer is: Deploy the job on AWS Batch using Amazon EC2 with appropriate compute resources.

AWS Batch is designed to efficiently run batch workloads by dynamically provisioning optimal compute resources. It allows the job to run on EC2 instances that meet the required specificationssuch as 64 vCPUs and 512 GiB memorywhile handling job queuing, instance provisioning, and scaling automatically. This dramatically reduces the operational overhead involved in infrastructure setup and maintenance. Because the task is CPU intensive and requires large resources for short bursts, AWS Batch is ideal for matching performance requirements while remaining cost-effective and low-maintenance.

The option to run the batch workload using AWS Lambda with automatic horizontal scaling is not viable because Lambda has hard limits on memory (up to 10 GiB) and execution time (15 minutes max). Furthermore, it's not designed for high-CPU, long-running workloads, and would not scale to meet the needs of the described job.

The suggestion to provision the workload in Amazon Lightsail and enable Auto Scaling for compute capacity is not a good fit. Lightsail is aimed at simpler applications with basic compute needs and lacks the flexibility and scalability to handle a large, CPU-bound batch job that requires high-performance compute instances. Auto Scaling in Lightsail is also limited compared to EC2-based services.

The choice to containerize the job and deploy it with Amazon ECS on Fargate for serverless orchestration offers some benefits in terms of reducing server management, but Fargate currently supports up to 16 vCPUs and 120 GiB memory per task, which is insufficient for the job's resource requirements. Scaling out would not meet the constraint of completing the job within 15 minutes, and splitting the job would require significant refactoring of the code.

This question is important because it challenges the candidate to match workload requirementssuch as CPU, memory, duration, and job schedulingto the most appropriate AWS compute service. It's easy to be tempted by 'serverless' options like Lambda or Fargate, but they are not always suitable for large, compute-heavy tasks. Recognizing the limitations of those services and understanding when to fall back on EC2-backed solutions like AWS Batch is key.

The trick to answering this type of question is to carefully evaluate performance needsespecially specific resource specs like vCPU and memoryand immediately eliminate options that cant meet those minimum thresholds. Pay close attention to wording like least operational overhead or minimal changes to distinguish between fully managed services and those requiring hands-on instance management.

Full AWS Practitioner Certification Question