Local Uploads for Cloudflare R2 – Global Upload Performance Boost

Local Uploads, launched in open beta in February 2026, write object data to a storage node near the client before asynchronously replicating it to the bucket’s primary region. This approach delivers immediate accessibility, strong consistency, and up to 75 % lower latency for cross‑regional uploads, making R2 truly global for both reads and writes.

How Local Uploads Work

When a client initiates an upload, the R2 Gateway Worker receives the request at the edge. If Local Uploads are enabled, the data is first stored in a regional storage cluster close to the client. Simultaneously, metadata is recorded in the Durable Object Metadata Service of the bucket’s home region. After the initial write, a background replication task copies the object to the bucket’s persistent storage, preserving the object storage model and strong consistency guarantees.

Standard Upload Flow

In the classic path, the client streams bytes to the edge, which encrypts and forwards them directly to the bucket’s region. Once the data lands, the metadata service updates the object’s index, and a 200 response is returned. Long‑distance transmission can increase latency and variability, especially for geographically dispersed users.

Local Upload Flow

With Local Uploads active, the edge node writes the payload to a nearby storage node. The metadata update still targets the bucket’s home region, but the client receives a success response as soon as the local write completes. The subsequent replication task runs asynchronously, ensuring the object becomes available in the primary region without additional client wait time.

Replication Tasks and Queues

Replication is orchestrated via Cloudflare Queues, which throttle task rates and provide retry and dead‑letter handling. Each replication job records a pending replica key, a timestamped marker, and a plan detailing source and destination locations, mode, and priority. This design balances cross‑regional bandwidth costs against latency goals.

Performance Impact and Use Cases

Beta testing showed median Time‑to‑Last‑Byte (TTLB) dropping from ~2 seconds to ~500 ms for uploads from North America to an Asia‑Pacific bucket. Applications that stream media, collect telemetry, or serve globally distributed users benefit from faster write paths without re‑architecting bucket locations. For detailed benchmark data, see Cloudflare’s 2026 product announcements and regional performance analyses in the regional performance benchmarks.

Ideal Workloads

Workloads with high write intensity across multiple continents—such as user‑generated video uploads, IoT telemetry pipelines, and global backup services—see the greatest latency reductions. Local Uploads also support scenarios where bucket relocation is impractical but write performance remains critical.

Limitations

Local Uploads are unavailable for jurisdiction‑restricted buckets (e.g., EU‑only or FedRAMP‑compliant containers) because regulatory constraints prohibit cross‑border data placement during the initial write.