SAP-C02 Question 628
Single answerA company is running a microservices-based application using Amazon Elastic Kubernetes Service (Amazon EKS) with managed node groups. The application requires dynamic scaling based on CPU utilization and needs to minimize operational overhead for managing the infrastructure. However, the company has observed that the nodes in the managed node group are sometimes underutilized, leading to unnecessary costs. What is the best solution to optimize cost and ensure scalability while adhering to the operational requirements?
- A
Use Amazon EKS Fargate profiles to run the pods, eliminating the need for managing nodes.
- B
Enable the Kubernetes Cluster Autoscaler with the managed node groups to scale the nodes dynamically based on pod requirements.
- C
Switch to using Amazon Elastic Container Service (Amazon ECS) with AWS Fargate to replace the Amazon EKS cluster for better cost management.
- D
Manually adjust the number of nodes in the managed node group during periods of high or low utilization.
Show answer and explanation
Correct answer: A
Explanation
Amazon EKS Fargate profiles enable serverless compute for pods, aligning with the company’s requirements to minimize operational overhead and optimize costs. Since Fargate abstracts the underlying infrastructure, the company does not need to manage EC2 instances, and they only pay for the resources consumed by the pods. This makes it the best solution compared to alternatives like Cluster Autoscaler or manual scaling, which still involve managing nodes or are operationally intensive.
- A. Correct.
Using Amazon EKS Fargate profiles allows pods to run on serverless compute, which eliminates the need to manage underlying nodes. This approach ensures that you only pay for the resources used by your pods, leading to cost optimization and reduced operational overhead.
- B. Incorrect.
The Kubernetes Cluster Autoscaler can dynamically scale the nodes in the managed node group. However, this still requires managing the underlying EC2 instances, which doesn’t align with the goal of minimizing operational overhead.
- C. Incorrect.
Switching to Amazon ECS with AWS Fargate might be a viable option, but it would require significant architectural changes to migrate from Amazon EKS to Amazon ECS. This is not the best solution for the given scenario.
- D. Incorrect.
Manually adjusting the number of nodes is operationally intensive and does not meet the requirement of minimizing operational overhead. It also introduces latency in scaling and is prone to human error.