SAP-C02 Practice Question 9: S3 Storage Cost Unpredictable Access
AWS Certified Solutions Architect – Professional (SAP-C02) · Test 1 · question 9 of 10
A media analytics platform stores 4 PB of rendered video segments and derived metadata files in a single Amazon S3 bucket, all in the S3 Standard storage class. Objects range from 1 MB to 5 GB. Analysis of access logs shows that most objects are read heavily for the first few weeks, then go quiet, but any object can be pulled back into a customer campaign months later, and when that happens the reads are heavy and sustained again. There is no reliable way to predict which objects will be revived.
Finance wants the storage bill reduced. The engineering constraints are that every object must remain retrievable in milliseconds by the existing application, which issues plain GetObject calls and cannot be changed to handle a restore step; the data must keep the durability and multi-Availability Zone resilience of S3 Standard; and the team does not want to build or run anything that inspects access patterns itself.
What should the solutions architect recommend?
Show answer and explanation
Answer:
S3 Intelligent-Tiering is the storage class AWS recommends for data with unknown, changing, or unpredictable access patterns. For a small monthly per-object monitoring and automation charge it moves objects that have not been accessed into the Infrequent Access tier and then the Archive Instant Access tier, and moves them straight back to the Frequent Access tier on the next read. All three of those tiers are low-latency, high-throughput tiers, so GetObject continues to return in milliseconds and the application needs no restore logic. Critically, there are no retrieval charges in S3 Intelligent-Tiering, which is what makes it right for objects that unpredictably come back to life; the revival is free. The optional Archive Access and Deep Archive Access tiers are asynchronous and must be left deactivated to preserve millisecond access. Intelligent-Tiering stores data across multiple Availability Zones with the same durability as S3 Standard, and S3 Storage Lens gives visibility into the savings without anyone writing analysis code.
Option A breaks the millisecond requirement outright: S3 Glacier Flexible Retrieval requires an explicit restore that takes minutes to hours, and the application cannot issue one. Even before 90 days, S3 Standard-IA charges a per-GB retrieval fee, so the unpredictable heavy re-reads described in the stem could cost more than the storage saved.
Option B fails the resilience requirement. S3 One Zone-IA stores data in a single Availability Zone, so it is not a substitute for S3 Standard's multi-AZ resilience for primary data, and it carries the same per-GB retrieval fees on revived objects.
Option D reimplements Intelligent-Tiering by hand, and badly. It is exactly the access-pattern analysis the team said it would not run, CloudTrail data events for a 4 PB bucket are expensive in their own right, the nightly cadence reacts slowly, and copying objects between classes incurs transition request charges and rewrites object metadata.
Reference: Managing storage costs with Amazon S3 Intelligent-Tiering