Back to roadmap
Module 7 · NoSQL, Search, Graph, ObjectDay 06625 min

Object Storage

Cheap, durable, infinite — your bucket of bytes.

← PreviousNext →
Day 066

Object Storage

Client
client
API
service
S3
datastore
Pipeline
service
Signal path
Direct-to-S3 upload via presigned URL
Client
client
presign
API
service
Client
client
upload
S3
datastore
S3
datastore
event
Pipeline
service
Memory hook

Object Storage: cheap, durable, infinite

Mental model

match the datastore to the access pattern

Design lens

Object stores aren't filesystems — no rename, eventual list.

Recall anchors
DurabilityClassesPatterns

Why it matters

Object stores (S3, GCS, Azure Blob) hold opaque blobs by key with rich metadata. They're the cheapest durable storage at scale and the natural home for media, backups, logs, and data lakes.

Deep dive

11 9s of durability via cross-AZ replication.

Storage classes: standard, infrequent access, archival.

Use multipart upload for large files; presigned URLs let clients upload/download without going through your servers.

Demo / scenario

Direct mobile upload of 5 GB videos.

  1. Server issues presigned multipart URL.
  2. Client uploads directly to S3.
  3. S3 emits event → ingest pipeline.
  4. Original moves to IA after 30 days, archive after 1y.

Tradeoffs

  • Object stores aren't filesystems — no rename, eventual list.
  • Per-object metadata bloats with too many.
  • Lifecycle policies must match access patterns.

Diagram

presignuploadevent
Client
API
S3
Pipeline
Direct-to-S3 upload via presigned URL.

Mind map

Check yourself

Loading quiz…

Sources & further reading

← PreviousNext →