A web application allows users to upload short videos. Videos are stored on EBS volumes and analyzed by custom recognition software. The site uses EC2 Auto Scaling for the web tier and a separate Auto Scaling group processing an SQS queue. The company wants to reduce operational overhead by using managed AWS services where possible and remove third-party software dependencies. Which re-architecture meets these goals?

The correct answer reduces operational overhead by leveraging fully managed services. Hosting the web application in Amazon S3 and storing videos there eliminates the need for EC2 instances and EBS volumes for the web tier and storage. S3 event notifications directly to SQS remove custom code for queueing. Processing the SQS queue with AWS Lambda is serverless, eliminating the Auto Scaling group for processing. Amazon Rekognition replaces the custom recognition software, removing third-party dependencies and management. Incorrect options: Running ECS containers still requires managing container infrastructure. Spot Instances for the SQS processing group don't fully eliminate operational overhead compared to serverless Lambda. Amazon EFS is a managed file system, but mounting it to EC2 instances for the web tier still involves EC2 management. While Lambda for SQS processing is good, EFS for video storage isn't as fully managed as S3 for static web hosting and video storage. Elastic Beanstalk simplifies deployment but still relies on EC2 instances and an Auto Scaling group, which is less serverless than the chosen solution.