Question confusion around S3 Intelligent-Tiering and Glacier Pricing

[mddreeves]

I am currently preparing to re-certify my SA Pro and am taking your practice quizzes in review mode. The below question has answer “A” as the answer, but I would argue it would be “C”. Since the data gets moved from S3 into FsX every 30 days, even though it is in Intelligent tiering, it will never move out of S3-Standard. 30 days of S3 standard is slightly more expensive than 30 days of Glacier instant retrieval, even with the retrieval charge. Am I wrong?

31. Question

Category: CSAP – Continuous Improvement for Existing Solutions

A company is running a financial modeling application on the AWS cloud. The application tier runs on an Auto Scaling group of Amazon EC2 instances. A separate EC2 cluster with a fixed number of instances is hosting the 200 TB of financial data in a shared file system. The application reads and processes the data on the shared filesystem to generate an overall financial report, which takes about 72 hours to complete. This whole process only needs to run at the end of each month, but the storage tier instances are running continuously to retain all the data in the shared file system.

As the storage tier takes up a large percentage of operational costs, the management wants to reduce the cost of the storage tier while maintaining the high-performance access needed by the application during its 72-hour run.

Which of the following options should the solutions architect implement that will have the largest overall cost reduction?

A. For the data tier, create an Amazon S3 bucket and move the objects of the existing shared file system to it. Use S3 Intelligent-Tiering Storage class to save costs. Use lazy-loading on an Amazon FSx for Lustre filesystem to import the contents of the S3 bucket. Use this filesystem as shared storage for the application tier EC2 instances for the duration of the job and delete it once the job is completed.

B. For the data tier, create an Amazon EFS filesystem and move the objects of the existing shared file system to it. Since the data will only be used once a month, use EFS Standard–Infrequent Access (IA) class to save costs. Mount this filesystem as shared storage for the application tier EC2 instance for the duration of the job.

C. For the data tier, create an Amazon S3 bucket and move the objects of the existing shared file system to it. Use S3 Glacier Instant Retrieval Storage class to save costs. Use lazy-loading on an Amazon FSx for Lustre filesystem to import the contents of the S3 bucket. Use this filesystem as shared storage for the application tier EC2 instances for the duration of the job and delete it once the job is completed.

D. For the data tier, create a large EBS volume with Multi-Attach enabled. Move the objects of the existing shared file system to it. This will save cost since only 1 EBS volume needs to be mounted on all application tier EC2 instances for the duration of the job. The EBS volume will be retained once the job is completed.

[JR-TutorialsDojo]

Hello mddreeves,

Thank you for the feedback.

The Glacier Instant Retrieval option is not the most cost-effective choice due to its minimum storage duration of 90 days. Since the data is only needed for a 72-hour job each month, it may be deleted or overwritten afterward. However, you would still incur charges for the full 90 days, even if you only use the data for 30 days. This can result in significantly higher cumulative costs, especially at a scale of 200 TB.

https://docs.aws.amazon.com/AmazonS3/latest/userguide/glacier-storage-classes.html#GIR

I hope this helps! Let us know if you need further assistance.

Regards,
JR @ Tutorials Dojo