AWS Certified Solutions Architect – Professional — Question 127
A company wants to manage the costs associated with a group of 20 applications that are infrequently used, but are still business-critical, by migrating to AWS.
The applications are a mix of Java and Node.js spread across different instance clusters. The company wants to minimize costs while standardizing by using a single deployment methodology. Most of the applications are part of month-end processing routines with a small number of concurrent users, but they are occasionally run at other times. Average application memory consumption is less than 1 GB, though some applications use as much as 2.5 GB of memory during peak processing. The most important application in the group is a billing report written in Java that accesses multiple data sources and often for several hours.
Which is the MOST cost-effective solution?
Answer options
- A. Deploy a separate AWS Lambda function for each application. Use AWS CloudTrail logs and Amazon CloudWatch alarms to verify completion of critical jobs.
- B. Deploy Amazon ECS containers on Amazon EC2 with Auto Scaling configured for memory utilization of 75%. Deploy an ECS task for each application being migrated with ECS task scaling. Monitor services and hosts by using Amazon CloudWatch.
- C. Deploy AWS Elastic Beanstalk for each application with Auto Scaling to ensure that all requests have sufficient resources. Monitor each AWS Elastic Beanstalk deployment by using CloudWatch alarms.
- D. Deploy a new Amazon EC2 instance cluster that co-hosts all applications by using EC2 Auto Scaling and Application Load Balancers. Scale cluster size based on a custom metric set on instance memory utilization. Purchase 3-year Reserved Instance reservations equal to the GroupMaxSize parameter of the Auto Scaling group.
Correct answer: B
Explanation
The correct answer is B because deploying Amazon ECS containers allows for effective resource management and cost efficiency by scaling based on memory utilization. Option A is not suitable as Lambda may not handle the memory requirements effectively for all applications. Option C, while useful, may incur higher costs due to individual Elastic Beanstalk environments. Option D involves a more complex setup with potential for higher operational costs due to the need for reserved instances and load balancing.