Covers the three universal storage types (block, file, object), S3 fundamentals and storage classes, the Snow family, the full catalog of AWS storage services, and hands-on demos of S3, EBS, and EFS.

The Three Types of Storage

TypeAWS serviceHow data is storedAccessWhen to use
BlockElastic Block Store (EBS)Data split into evenly sized blocks, directly accessed by the OSSingle write volumeYou need a virtual drive attached to your VM 5:12:20
FileElastic File System (EFS)File stored with data and metadataMultiple connections via network share (NFS/SMB); multiple reads, writes lock the fileMultiple users/VMs need to access the same drive 5:13:09
ObjectSimple Storage Service (S3)Object stored with data, metadata, and a unique IDHTTPS/API; multiple reads and writes, no locksUpload files without worrying about underlying infrastructure; not intended for high IOPS 5:14:02
  • Choice depends on how fast reads/writes must be, how many clients connect simultaneously, and whether the storage must be mounted to a VM 5:14:39

S3 Fundamentals

  • S3 is AWS's flagship storage service and the second AWS service ever launched; it provides basically unlimited object storage 5:14:58
  • Objects consist of: key (name), value (sequence of bytes), version ID (when versioning is enabled), and metadata. Buckets hold objects (and folders which hold objects) 5:15:49
  • S3 is a universal namespace — bucket names must be globally unique, like domain names 5:16:13
  • Exam fact: an individual object can be 0 bytes up to 5 TB (yes, a zero-byte file is valid) 5:16:13

S3 Storage Classes

Classes trade retrieval time, accessibility, and durability for cheaper storage 5:16:38:

  • S3 Standard — default tier: 99.99% availability, 11 nines durability, replicated across 3 AZs 5:17:13
  • S3 Intelligent-Tiering — ML analyzes usage and moves objects to the most cost-effective tier with no performance impact or overhead 5:17:41
  • S3 Standard-IA (Infrequent Access) — as fast as Standard, cheaper if accessed less than once a month, but adds a retrieval fee (frequent access ends up costing MORE than Standard) 5:17:41
  • S3 One Zone-IA — single AZ: as fast as Standard, cheaper, lower availability; caveat: data can be destroyed if that AZ suffers a catastrophe 5:18:06
  • S3 Glacier — long-term cold storage; retrieval takes minutes to hours but very cheap. S3 Glacier Deep Archive — lowest cost class; retrieval takes 12 hours 5:18:28
  • S3 Outposts has its own storage class and doesn't fit the linear cheaper-tier progression; Glacier is basically its own service but part of S3 5:19:00

AWS Snow Family

  • Storage and compute devices to physically move data in/out of the cloud when internet transfer is too slow, difficult, or costly 5:19:27
  • Snowcone: two sizes — 8 TB and 14 TB usable. Snowball Edge: storage optimized 80 TB usable, or compute optimized (~39.5 TB) with many vCPUs and increased memory for edge computing before shipping to AWS. Snowmobile: up to 100 PB per trailer 5:19:53
  • All data lands in Amazon S3. AWS markets Snowmobile as "move exabytes" only because you can order multiple trailers — a single one is 100 PB 5:20:45
  • Original Snowball is likely no longer available — just Snowball Edge (briefcase-sized, 50–80 TB) 5:22:40

Storage Services Overview

  • S3 — serverless object storage: unlimited files, very large files, pay for what you store, no file system/disk management. S3 Glacier — cold storage on previous-generation HDDs for low cost; highly secure and durable 5:21:14
  • EBS — persistent block storage, a "virtual hard drive in the cloud" attached to EC2; drive types: SSD, IOPS SSD, throughput HDD, cold HDD 5:21:41
  • EFS — cloud-native NFS file system, mountable to multiple EC2 instances at the same time. Storage Gateway — hybrid cloud storage extending on-premise storage to the cloud, with three offerings: File Gateway (extend local storage to S3), Volume Gateway (caches local drives to S3 for continuous backup), Tape Gateway (virtual tapes for cost-effective long-term backup) 5:22:07
  • AWS Backup — fully managed, centralizes and automates backups across services (EC2, EBS, RDS, DynamoDB, EFS, Storage Gateway) via backup plans. CloudEndure Disaster Recovery — continuously replicates machines to a low-cost staging area in a target account/region for fast recovery from data center failures. Amazon FSx — feature-rich high-performance file system: FSx for Windows File Server (SMB protocol, mount to Windows servers) and FSx for Lustre (Linux) 5:23:05

Demo: S3 in Practice

  • Bucket types: general purpose (recommended for basically everything) vs directory (low latency, works with S3 Express One Zone) 5:25:07; a newer account regional namespace option adds a suffix so bucket names no longer conflict globally 5:25:31
  • ACLs are legacy and not recommended — use bucket policies instead. Block Public Access is on by default; you must explicitly remove the safeguard (and own the risk of data leaks) to make anything public 5:26:22
  • Versioning keeps files recoverable (an attacker must delete every version); encryption options: SSE managed by AWS, your own key managed by AWS, or dual-layer server-side encryption (bucket key now on by default) 5:27:35
  • Object Lock (requires versioning) prevents overwrite/delete for a fixed time or indefinitely — strong protection against data loss or account compromise 5:27:54
  • Pre-signed URLs share a downloadable link that expires (e.g., 5 minutes) — the standard way to share private objects 5:31:45
  • Objects have a URL but return Access Denied until a bucket policy allows s3:GetObject; the policy needs a Principal (e.g., * = anyone) and a Resource ARN — scoping it to a specific object is safer than allowing the whole bucket 5:33:40, demonstrated: only the whitelisted file was readable, a second upload stayed denied 5:37:00
  • Lifecycle rules move files between tiers over time (e.g., Standard → Glacier after a period) or permanently delete non-current versions after N days 5:37:23
  • To delete a bucket you must empty it first 5:39:30

Demo: EBS

  • EBS lives under the EC2 console (Volumes, Snapshots, lifecycle management); snapshots are saved versions of volumes 5:40:12
  • Volumes are usually created at instance launch (Configure Storage), not separately; free tier allows up to 30 GB GP3 5:40:38
  • GP3 vs GP2 availability varies by region/AZ; newer is better 5:41:06
  • AWS does not encrypt EBS volumes by default (other providers do) — encrypt at launch, because enabling it later requires snapshotting and relaunching 5:41:34; some volume types cannot be a root volume 5:42:23

Demo: EFS

  • EFS connects multiple VMs (or even your local computer) to the same shared drive 5:42:36
  • Options: regional vs one-zone (one-zone for low latency), backups, lifecycle management, encryption at rest, and throughput modes — elastic (pay for what you use, most cost-effective) vs provisioned 5:43:28
  • EFS creates mount targets in every availability zone so you can mount from anywhere in the VPC; VMs must generally be in the same VPC 5:44:37; an access point is needed to utilize it, and the EFS mount helper greatly simplifies mounting 5:45:48
  • No volume size is chosen — it is basically serverless storage that scales (and bills) on what you use 5:46:11
  • Proof of persistence: a file created under /mnt/efs from one EC2 instance survived instance teardown and was readable from a brand-new instance — the file system lives independently of EC2; multiple instances can connect simultaneously 5:49:55
  • Cleanup order: delete access points first, then the file system 5:52:03