A company manages an application that stores logs in Amazon CloudWatch Logs.

To archive CloudWatch Logs to S3 with long-term retention, you need:

a streaming mechanism to move data from CloudWatch Logs to S3, and

an S3 Lifecycle policy to handle tiering and expiration.

Option B uses a CloudWatch Logs subscription filter that targets Amazon Data Firehose. Firehose provides managed, scalable streaming from CloudWatch Logs to S3 with optional buffering and transformation. This is the AWS-recommended pattern for exporting continuous log streams with minimal operational overhead. Option C is not valid because CloudWatch Logs subscription filters cannot directly target S3; they must target Kinesis, Firehose, or Lambda.

Once logs are in S3, the company wants to keep them rarely accessed after 90 days and retained for 10 years. Option D configures an S3 lifecycle policy to transition logs to S3 Glacier Instant Retrieval after 90 days, which is a low-cost archival tier with relatively fast access, and to expire (delete) logs after 3,650 days (10 years). This precisely matches the retention requirement.

Option E uses Reduced Redundancy, which is a legacy storage class and not optimized for long-term archival. Therefore, the correct combination is B and D.