AWS Certified Solutions Architect Professional SAP-C02 Practice Question

The correct answer is to use Amazon ECS on EC2 with G4dn Spot Instances within a Capacity Provider. This solution meets all the specified requirements. Amazon ECS is a fully managed container orchestration service that simplifies container deployment and management, which aligns with the goal of minimizing operational overhead for a team with limited Kubernetes experience. The EC2 launch type for ECS allows for the use of GPU-accelerated instance types, such as the G4dn instances, fulfilling the mandatory GPU requirement. To address the spiky workload and optimize costs, an ECS Capacity Provider configured with EC2 Spot Instances is the ideal approach. Spot Instances offer significant cost savings for fault-tolerant, long-running batch jobs, and Capacity Providers automate the scaling of these instances based on workload demand.

Incorrect Answers:

• Amazon EKS with managed node groups using G4dn EC2 instances: While Amazon EKS also supports GPU instances, it is based on Kubernetes and generally introduces a higher level of operational complexity and a steeper learning curve compared to Amazon ECS. Given the team's limited Kubernetes expertise, ECS is the more appropriate choice to minimize operational burden.

• AWS Fargate with a specified vCPU, memory, and ephemeral storage configuration: This option is incorrect because AWS Fargate, the serverless compute engine for containers, does not support GPU instances. The requirement for GPU acceleration is a hard constraint that Fargate cannot meet.

• AWS Batch using a compute environment configured with AWS Fargate: This option is incorrect for the same reason as the Fargate-only option. While AWS Batch is an excellent service for managing batch computing workloads, the specified compute environment is AWS Fargate, which lacks GPU support. To use GPUs with AWS Batch, an EC2-based compute environment would be required.