AWS provides multiple storage services for different workloads, each built with specific performance, durability, and access patterns. The three core storage services—Amazon S3, Amazon EBS, and Amazon EFS—support various sub-types to optimize cost, performance, and availability. This guide explains their architectures and dives deeper into every storage class, volume type, and performance mode.
1. Amazon S3 (Simple Storage Service)
Amazon S3 is a fully-managed, multi-AZ object storage service built for massive scalability, strong consistency, and 11 9’s durability. It is ideal for unstructured data, backups, static assets, and large analytics workloads.
Technical Architecture
- Object-based storage with metadata and unique identifiers.
- Strong read-after-write consistency for PUT and DELETE operations.
- Multi-AZ replication ensures extremely high durability.
- Designed for parallel access and high request concurrency.
S3 Storage Classes Explained
S3 Standard
- High durability and low latency.
- Ideal for frequently accessed hot data.
- Used for websites, applications, and daily analytics.
S3 Standard-IA (Infrequent Access)
- Lower cost than Standard.
- Minimum 30-day retention.
- Used for backups or data accessed once per week/month.
S3 One Zone-IA
- Stored in a single AZ.
- Lower durability but cheaper than Standard-IA.
- Used for re-creatable data (temporary datasets, scratch data).
S3 Intelligent-Tiering
- Automatically moves objects between frequent and infrequent tiers.
- No operational overhead.
- Perfect for unknown or changing access patterns.
S3 Glacier Instant Retrieval
- Low-cost archival with millisecond retrieval.
- Ideal for medical documents, compliance archives.
S3 Glacier Flexible Retrieval
- Hours retrieval time.
- Cheaper than Instant.
- Used for long-term archives with occasional restores.
S3 Glacier Deep Archive
- Cheapest S3 tier.
- 12–48 hours retrieval.
- Used for 7–10 year compliance storage.
S3 Lifecycle Rules
- Automatically transition data to cheaper classes.
- Automatically delete expired objects.
2. Amazon EBS (Elastic Block Store)
Amazon EBS is EC2-attached block storage designed for extremely low-latency and high throughput workloads. It behaves like a durable virtual SSD or HDD.
Technical Architecture
- Network-attached block storage replicated within a single AZ.
- Consistent sub-millisecond latency for SSD volumes.
- Snapshots stored in Amazon S3 for backup or restore.
- Encryption-at-rest and in-transit supported via KMS.
EBS Volume Types (SSD & HDD)
SSD-Based Volumes
• gp3 (General Purpose SSD)
- Baseline 3,000 IOPS & configurable up to 16,000 IOPS.
- Throughput up to 1,000 MB/s.
- Ideal for web servers, development environments.
• io2 / io2 Block Express (Provisioned IOPS SSD)
- Up to 256,000 IOPS (Block Express).
- Sub-millisecond latency.
- Designed for high-end databases like Oracle, SAP HANA.
HDD-Based Volumes
• st1 (Throughput-Optimized HDD)
- Optimized for sequential I/O.
- Used for big data, streaming, log processing.
• sc1 (Cold HDD)
- Lowest EBS cost.
- Used for archiving and infrequently accessed data.
EBS Multi-Attach
- Supported on io2 volumes.
- Allows up to 16 EC2 instances to share the volume concurrently.
- Used for clustered applications requiring shared block storage.
EBS Snapshots
- Incremental backups; only the changed blocks stored.
- Cross-Region and Cross-Account replication available.
- Snapshots are stored durably in S3.
3. Amazon EFS (Elastic File System)
Amazon EFS offers a scalable, elastic NFS file system accessible by multiple EC2 instances across multiple Availability Zones.
Technical Architecture
- NFSv4.1/4.0 protocol support.
- Distributed multi-AZ file system with automatic scaling.
- Highly available and durable.
- No provisioning needed—pays only for used data.
EFS Storage Classes
EFS Standard
- Multi-AZ file system for general workloads.
- Low-latency read/write operations.
EFS Standard-IA (Infrequent Access)
- Cheaper storage for data accessed occasionally.
- Automatically managed by lifecycle policies.
EFS One Zone
- Stored in one AZ only.
- Lower cost than Standard.
- Used for dev/test workloads.
EFS One Zone-IA
- Cheapest EFS tier.
- Used for non-production workloads with low access frequency.
EFS Performance Modes
General Purpose (Default)
- Low latency.
- Used for web apps, content management, home directories.
Max I/O
- Higher throughput for large-scale workloads.
- Slightly higher latency.
- Used for big data, analytics clusters, and media processing.
EFS Throughput Modes
- Bursting Throughput: Default mode; scales with data stored.
- Provisioned Throughput: Specify exact MB/s required for high-performance workloads.
Comparison Table: S3 vs EBS vs EFS
| Feature | S3 | EBS | EFS |
|---|---|---|---|
| Storage Type | Object | Block | File (NFS) |
| Durability | Multi-AZ (11 9’s) | Single AZ | Multi-AZ |
| Access | REST / HTTP(S) | Attach to EC2 | Mount on multiple EC2 |
| Use Case | Analytics, backup, static content | Databases, OS, low-latency apps | Shared file systems, container workloads |
| Performance | High throughput | High IOPS | Shared throughput |
| Capacity | Unlimited | Up to 64 TiB | Petabytes |
Choosing the Right AWS Storage Service
- Choose S3 for scalable object storage, backups, static content, and large datasets.
- Choose EBS for transactional systems, databases, and workloads requiring low-latency block-level storage.
- Choose EFS for shared-file storage accessed by multiple compute instances or containers.
A deep understanding of the sub-types—such as S3 storage classes, EBS volume types, and EFS performance modes—helps choose the optimal mix of cost and performance for any cloud workload.
Views: 27