Cloudflare R2

Use it when

• Egress is the cost that hurts: R2 charges nothing for data transfer out at any volume, which is what makes multi-engine and multi-cloud reads affordable.
• Data is served to the internet or into Workers, where the free-egress shape compounds.
• Existing S3 tooling should move over by changing an endpoint and credentials.
• An Iceberg lakehouse on the bucket appeals: R2 Data Catalog is a managed Iceberg REST catalog reachable from Spark, Snowflake, and PyIceberg.

Think twice when

• You need the long tail of S3 features; ACLs, versioning, tagging, replication rules, and per-object locks are absent.
• Data must live in a chosen region with cross-region replication; buckets are single-region with location hints and jurisdiction restrictions only.
• The catalog is the point today: R2 Data Catalog is public beta, worth evaluating but not yet depending on.
• Heavy compute sits in another cloud's region anyway, where that provider's store keeps traffic in-region and free.

How it runs

Fully managed with eleven-nines durability from replication and erasure coding within a region. Billing is per GB-month plus Class A and Class B operations, nothing for egress. Events flow to Queues consumed by Workers, and a standing free tier covers 10 GB and millions of operations monthly.