aws-storage

Match the access pattern to the storage category: S3 for object API access at scale, EFS for shared Linux NFS file access, and FSx variants for specialized needs like Lustre HPC performance, ONTAP multi-protocol enterprise features, OpenZFS low latency, or Windows SMB. The skill compares candidates against capacity, performance, durability, and cost requirements.

S3 Vectors provides vector storage and similarity search on S3 with native support for storing and querying embeddings cost-effectively. It suits RAG pipelines, semantic search, recommendation systems, and AI agent memory workloads.

Yes, EFS is mountable by Lambda, Fargate, EC2, ECS, and EKS, making it suitable for serverless persistent storage. S3 Files can also be mounted from Lambda and Fargate when data already lives in an S3 bucket.

FSx for NetApp ONTAP supports SnapMirror replication for NetApp sources, while AWS DataSync handles migrations from non-NetApp NAS systems and preserves NTFS ACLs when moving to FSx for Windows File Server. The skill routes to migration references based on source system and protocol.

S3 returns 503 SlowDown when request rates exceed prefix partitioning limits. Spread requests across prefixes and retry with exponential backoff and jitter; for latency-sensitive high-request workloads, consider S3 Express One Zone directory buckets instead.

The skill does not cover SQL query engines (Athena, Redshift, EMR), ETL services (Glue), streaming platforms (Kafka, MSK, Kinesis), or managed databases (RDS, Aurora, DynamoDB). For deep cost analysis beyond storage selection, it routes to the billing-and-cost-management skill.