Services at a glance

Amazon S3 (Simple Storage Service)

Deep dive: AWS S3 Deep Dive (storage classes, lifecycle, versioning and Object Lock, encryption, access control, replication, performance).

Object storage with virtually unlimited capacity. Objects (up to 5 TB) live in buckets; bucket names are globally unique, buckets are regional. Designed for 99.999999999% (11 nines) durability.

• Object = key (the "path"), data, metadata, version id
• Upload > 5 GB must use multipart upload (recommended > 100 MB)
• Strong read-after-write consistency

Storage classes

Lifecycle rules transition or expire objects by age.

Features

• Versioning: keep every version, protects against overwrite/delete; MFA delete
• Replication: CRR (cross-region) and SRR (same-region), needs versioning
• Static website hosting
• Event notifications to Lambda, SQS, SNS, EventBridge
• Pre-signed URLs: time-limited access to a private object
• Object Lock (WORM), Glacier Vault Lock
• S3 Select / Athena: query data in place
• Transfer Acceleration: faster uploads via edge locations

Security

• Buckets are private by default; Block Public Access at account/bucket level
• Access via IAM policies, bucket policies, ACLs (legacy), access points
• Encryption: SSE-S3 (default), SSE-KMS, SSE-C, client-side
• Access logs and CloudTrail data events

$ aws s3 mb s3://my-unique-bucket-12345
$ aws s3 cp ./report.pdf s3://my-unique-bucket-12345/reports/
$ aws s3 sync ./site s3://my-unique-bucket-12345/ --delete
$ aws s3 presign s3://my-unique-bucket-12345/reports/report.pdf --expires-in 3600

Block and File Storage

Amazon EBS

Network-attached block volumes for EC2, AZ-scoped, persist independent of the instance.

• Snapshots are incremental, stored in S3, copy across regions/accounts; restore to a new volume
• Encryption with KMS, Multi-Attach for io1/io2
• Instance store: physically attached, very fast, lost when the instance stops or fails

Amazon EFS

Managed NFS file system, elastic, shared by many instances across multiple AZs. Linux only. Classes: Standard, IA.

Amazon FSx

Managed third-party file systems: FSx for Windows (SMB, AD), Lustre (HPC), NetApp ONTAP, OpenZFS.

Hybrid and migration

• Storage Gateway: File / Volume / Tape gateway connecting on-prem to cloud storage
• DataSync: automated online transfer
• Snow Family: Snowcone / Snowball / Snowmobile for offline transfer

Relational Databases

Amazon RDS

Managed relational engines: MySQL, PostgreSQL, MariaDB, Oracle, SQL Server, Db2. AWS handles patching, backups and failover; you cannot SSH to the host.

• Multi-AZ: synchronous standby in another AZ for high availability / failover (not for scaling reads)
• Read replicas: asynchronous, up to 15, for read scaling, can be cross-region, can be promoted
• Automated backups (point-in-time restore, 1 to 35 days) and manual snapshots
• RDS Proxy: pools connections, vital for Lambda
• Storage auto scaling, encryption at rest with KMS, IAM database authentication

Amazon Aurora

Cloud-native MySQL/PostgreSQL-compatible engine. Up to 5x MySQL / 3x PostgreSQL throughput claimed.

• Storage replicated 6 copies across 3 AZs, grows automatically to 128 TB
• Up to 15 read replicas, sub-30 s failover
• Aurora Serverless v2: capacity scales automatically
• Global Database: cross-region replication, <1 s lag

Relational concepts are in the Relational db Concepts notes under Databases.

NoSQL and Other Databases

DynamoDB

Serverless key-value / document database with single-digit ms latency at any scale.

• Primary key: partition key, or partition + sort key
• Capacity modes: On-demand (pay per request) or Provisioned (RCU / WCU, with auto scaling)
• 1 RCU = one strongly consistent read of up to 4 KB per second (2 eventually consistent)
• 1 WCU = one write of up to 1 KB per second
• Indexes: GSI (different keys, any time) and LSI (same partition key, created at table creation)
• DynamoDB Streams feed Lambda; TTL expires items; Global Tables multi-region active-active
• PITR and on-demand backups; DAX in-memory cache for microsecond reads
• Prefer Query over Scan

$ aws dynamodb put-item --table-name Orders \
    --item '{"OrderId":{"S":"1001"},"Total":{"N":"99.5"}}'
$ aws dynamodb get-item --table-name Orders --key '{"OrderId":{"S":"1001"}}'

Other

Caching strategies

• Lazy loading (cache-aside): populate on a miss; can serve stale data
• Write-through: update cache on every write; no staleness, more writes

Backup and Disaster Recovery Aids

• AWS Backup: central policy-based backups across EBS, RDS, DynamoDB, EFS, S3 and more
• Snapshots and cross-region copies
• See 7. AWS Solutions Architecture for DR strategies