Teams often compare object storage providers by looking only at the cost per gigabyte stored. That is the wrong number for media platforms, download-heavy SaaS products, and public asset pipelines. In those systems, egress can dominate the bill long before raw storage does.
That is why storage architecture is really traffic architecture.
The Cost Model Most Teams Miss
A simple framing is more useful than a fancy spreadsheet:
monthly cost = stored data + request cost + transferred data
If a 5 TB asset library gets downloaded 80 TB per month, the transfer line matters far more than the bytes resting on disk. That is why services like Cloudflare R2 changed the conversation. They pushed teams to evaluate where data moves, not just where it sits.
Why This Changes Architecture Decisions
Once egress becomes material, teams start revisiting choices they treated as settled:
- where images and downloads are served from
- whether origin and CDN should be on the same vendor boundary
- whether private and public assets should live in different storage tiers
- whether multi-region replication is worth the transfer cost
That does not mean R2 is always the answer. It means S3 is not automatically the cheapest answer once usage becomes download-heavy.
The Trade-Offs Are Still Real
Moving storage vendors is not an accounting-only decision. You still need to check:
- ecosystem compatibility
- operational tooling
- migration effort
- latency to your compute layer
- compliance or residency requirements
A cheaper egress story can still be the wrong move if the rest of the platform becomes harder to operate.
Better Rule
Evaluate storage providers using actual request and transfer patterns from production or realistic estimates from logs. If the architecture serves a lot of bytes outward, egress belongs in the first conversation, not the postmortem after the bill arrives.
Further Reading