Examkin

Question

Difficulty: Very hardAWS Compute Services

A company is building a batch processing solution for genetic sequencing data. The application requires direct access to GPU hardware, a customized Linux kernel with specific compliance patches, and runs jobs that can take up to 4 hours to complete. The workload is highly unpredictable, and the company wants to minimize costs by ensuring resources are only active during processing. Which AWS compute option best meets these requirements?

  1. A
    AWS Lambda functions triggered by Amazon S3 events
  2. Amazon EC2 instances in an Auto Scaling group utilizing Spot InstancesAnswer
  3. C
    AWS Fargate tasks running on Amazon ECS
  4. D
    Amazon Lightsail container services with pre-configured stacks

Answer

Amazon EC2 instances in an Auto Scaling group utilizing Spot Instances
Amazon EC2 is the correct choice because it provides virtual machines where the customer has full root access, enabling the use of custom kernels and direct GPU mapping. Since the job runs for 4 hours, it exceeds Lambda's limits. EC2 Auto Scaling allows the group to scale to zero instances to minimize costs, and Spot Instances offer deep discounts suitable for batch jobs.

Step-by-Step Solution

1
Analyze the workload duration requirement.
The job takes up to 4 hours, which immediately rules out serverless functions like AWS Lambda due to their 15-minute execution limit.
Compute options must support long-running execution windows.
2
Evaluate operating system and hardware requirements.
The requirement for custom kernel compliance patches and direct GPU access rules out serverless container options like AWS Fargate, which abstract away the host OS and kernel management.
Administrative control over the OS kernel is a shared responsibility aspect that requires full virtual machine control (EC2).
3
Assess cost-optimization and scalability requirements.
Amazon EC2 Auto Scaling groups can scale to zero instances when no work is present, and utilizing Spot Instances matches the unpredictable, batch-processing nature of genetic sequencing at a highly discounted rate.
Minimizing idle costs requires dynamic scaling and leverage of excess AWS capacity pricing.

Key Concept

AWS Compute Service Selection based on compliance, hardware access, execution duration, and cost optimization
Rate this question

Topics

AWS Compute Services

More questions from this topic