Open Dropbox (File Sync) →Open Amazon S3 (Object Storage) →

Dropbox (File Sync) vs Amazon S3 (Object Storage)

Dropbox (File Sync)

Chunking, content-addressed dedup, multi-device sync.

Components (7)

Desktop Client
API Gateway
Metadata Service
Upload Service
Chunk Store
MySQL
Notification Service

Headline numbers

Sync ops / sec (avg)~50K
Raw bytes uploaded / day~350 TB (post-dedup)
Bandwidth at peak~30 Gbps

Amazon S3 (Object Storage)

Strongly consistent object storage: a metadata plane that locates objects, and a replicated (or erasure-coded) data plane spread across availability zones.

Components (9)

Client
API Service
Identity & Access Mgmt
Metadata Service
Data Routing Service
Placement Service
Primary Data Node · AZ1
Secondary · AZ2
Secondary · AZ3

Headline numbers

Erasure-coding overhead vs replication~50% less
Expected object losses / yr per 10B objects~0.1
Per-prefix QPS ceiling~5,500 GET / 3,500 PUT

Key differences

Only in Dropbox (File Sync)

Notification Service

In both

Client
Api Gateway
Service
Storage
Database

Only in Amazon S3 (Object Storage)

None.

Flow shape

Dropbox (File Sync) flows

Upload a modified file7 steps
Tiny edit to a 1GB file4 steps
⚠ Metadata Service is down3 steps

Amazon S3 (Object Storage) flows

PUT an object (replication)8 steps
GET an object5 steps
⚠ Alternative: erasure coding + AZ loss5 steps
⚠ Primary data node fails3 steps