The best design is to place an internet-facing Application Load Balancer in public subnets and place the EC2 instances in private subnets with no direct internet access. This keeps the application reachable from the internet while preventing direct internet connectivity to the instances. Because the clients use HTTP, an ALB is more appropriate than an NLB because ALB operates at Layer 7 and supports HTTP routing, health checks, host-based routing, and path-based routing. The EC2 instances should be part of an Auto Scaling group so that the application can scale based on traffic demand. Options that place the instances in public subnets violate the security requirement. The NLB options are less suitable for HTTP application-layer routing.

The correct answer isAbecause the application processes videos for5-30 minutes, must runmultiple jobs in parallel, and should have theleast operational overhead.Amazon ECS with AWS Fargateis a managed container solution that removes the need to provision and manage EC2 instances while still supporting longer- running parallel jobs. UsingAmazon SQS standard queuesdecouples the upload event from processing and provides a scalable buffer for incoming work.
With this design, Amazon S3 sends object-created notifications to the SQS queue whenever users upload videos. ECS services or tasks on Fargate can then consume messages from the queue and process videos independently. Auto scaling based onSQS queue depthensures the system increases task count when more videos arrive and decreases capacity when demand drops. This provides elasticity and efficient parallel processing with minimal infrastructure management.
Option B is incorrect because EC2-based scaling introduces more operational overhead than Fargate. Option C is incorrect because AWS Lambda has a maximum execution duration and is not appropriate for jobs that can run up to 30 minutes. Option D is also incorrect for the same reason; Lambda is not the best fit for this processing duration, and SNS does not provide the same durable queued work pattern as SQS for controlled parallel processing.
AWS best practices recommend usingevent-driven queues with containerized workersfor medium-duration processing jobs that need concurrency and minimal management. Therefore,S3 + SQS + ECS on Fargate with queue-based scalingis the best solution.

Amazon SQS provides durable, decoupled message storage for distributed systems. Using SQS as a job queue enables each EC2 instance to process a message independently. Scaling the Auto Scaling group based on the SQS queue length ensures parallelism and elasticity, aligning compute resources with workload volume.
Reference: AWS Documentation - Using Amazon SQS with Auto Scaling

AWS guidance for NAT Gateway recommends deploying "a NAT gateway in each Availability Zone and configure your routing to ensure that resources use the NAT gateway in the same Availability Zone." This provides "zone-independent architecture" and avoids cross-AZ data processing charges and single-AZ failures. Option B creates a single point of failure and incurs cross-AZ egress charges when private subnets in other AZs traverse a centralized NAT. NAT instances (C) are legacy, require manual scaling/failover
/patching, and are not recommended for production HA. Option D is not supported (NLB cannot front a NAT Gateway as a target). With steady, uniform load, per-AZ NAT Gateways deliver high availability with predictable cost; routing each private subnet to its local NAT Gateway maintains security (no inbound initiated connections) and resilience. This meets the requirement for cost-effective, secure outbound connectivity across multiple AZs while preserving availability.
References: VPC NAT Gateway documentation - Multi-AZ best practices and same-AZ routing; AWS Well- Architected Framework - Reliability and Cost Optimization (avoid single points of failure; minimize cross- AZ data transfer).

Amazon S3 Express One Zone provides single-digit millisecond latency and high throughput, making it ideal for ML workloads that require multiple compute nodes and fast access. It is also more cost-effective than traditional file or block storage for temporary, high-speed needs.
Reference: AWS Documentation - Amazon S3 Storage Classes