Data Lifecycle Management

• Pick the S3 storage class by retrieval speed first, then access frequency
• Use Standard-IA for data read a few times a quarter, not hot data
• One Zone-IA stores one AZ copy, so only use it for reproducible data
• Glacier Instant Retrieval gives millisecond reads for rarely-accessed archives
• Match the Glacier Flexible restore tier to how long you can wait
• Deep Archive is the cheapest class, with 12-to-48-hour restores
• Use Intelligent-Tiering when the access pattern is unpredictable
• All S3 classes give 11 nines durability except One Zone-IA
• Lifecycle transition actions move objects to cheaper classes by age
• A lifecycle rule cannot transition to IA before 30 days
• Lifecycle expiration actions delete objects at a set age
• Prune noncurrent versions with their own lifecycle actions
• S3 Versioning keeps every version and turns deletes into markers
• MFA Delete gates permanent version deletes and suspending versioning
• DynamoDB TTL auto-expires items on an epoch timestamp at no write cost
• Expired DynamoDB items still appear in reads until TTL removes them
• Use Redshift COPY for bulk parallel loads, not many INSERTs
• Use Redshift UNLOAD to export query results to S3 as Parquet
• Use S3 Replication for DR, low-latency reads, or data residency
• Use AWS Backup for centralized, policy-driven recovery across services
• S3 Lifecycle skips objects smaller than 128 KB by default
• When a Lifecycle object is eligible for both, expiration wins over transition
• Kinesis Data Streams retention runs 24 hours to 365 days, billed in tiers above 7 days
• Increasing Kinesis retention never brings back already-expired records
• Tier OpenSearch indexes hot to UltraWarm to cold with ISM policies

Unlock with Premium — includes all practice exams and the complete study guide.