A media company is designing a new application that ingests large video files uploaded by users. The files must be stored durably and then processed by a compute fleet that runs for several hours per file. The company wants to minimize compute costs and ensure that processing can resume if interrupted. Which architectural approach should a solutions architect recommend?
AWS Batch is designed for batch computing workloads and can run jobs for hours or days. It manages compute resources dynamically, which can reduce costs by using Spot Instances and scaling down when not needed. AWS Batch also supports job queues and retries, allowing processing to resume if interrupted, and integrates with S3 for input and output.
Why this answer
AWS Batch is purpose-built for batch computing, supporting long-running jobs and managed compute environments that can use Spot Instances to reduce costs. It provides job queues, retries, and dependency management, enabling processing to resume after interruptions. Storing files in Amazon S3 ensures durability and accessibility for the compute fleet.
This combination meets the requirements for cost minimization and resiliency.
Exam trap
The trap here is assuming that AWS Lambda or Fargate can handle multi-hour processing without considering their timeout limits or the lack of built-in batch orchestration.