TL;DR
Get the latest gadgets delivered free — and shop member deals
- Fast, free delivery on millions of items
- Access to Prime Big Deal Days deals on October 6–7
- Prime Video, Amazon Music and more included
Cloudflare has launched K2 in public beta, a serverless event-streaming service for storing ordered events and letting consumers read them independently. It is built on Cloudflare R2 and supports long-term retention, but the company reports about one second of producer latency at the 99th percentile in the initial release.
Cloudflare has launched K2 in public beta, a serverless event-streaming service on its Developer Platform that stores events in an ordered log so multiple consumers can read them independently. The company says K2 is built on R2 object storage and is intended to retain data through consumer outages, addressing a common problem in systems where producers and consumers must operate at the same pace.
With K2, applications send events to a stream, where the service stores them before consumers retrieve them. Cloudflare says readers can divide work among a group of consumers or receive copies of all messages, supporting use cases such as sending transaction events to both analytics and fraud-detection systems. The service is serverless and supports long-term retention, according to the company.
Cloudflare says K2 implements a partitioned, durable log on top of R2. Rather than append each event directly to an object, the service accumulates writes in memory on an edge service, then writes groups of events as segment files. The company says it uses R2 atomic operations to maintain ordering and strictly increasing offsets without a separate coordination service.
The design comes with a latency trade-off. Cloudflare reports that, in the initial release, waiting for a batch to accumulate and then writing it to object storage adds about one second of producer latency at the 99th percentile. That figure is Cloudflare’s reported measurement; the blog post does not give benchmark conditions or comparative results.
A Durable Buffer for Edge Workloads
K2 is aimed at systems that need to accept events even when downstream services are slow or unavailable. By separating producers from consumers, it can let applications continue sending data while readers catch up later. That may reduce the risk of losing accepted events during interruptions and allows separate services to process the same stream for different purposes.
The product also reflects a design choice shaped by Cloudflare’s infrastructure. The company says its edge network spans servers in more than 335 cities, but describes those machines as relatively ephemeral and available in smaller slices than a conventional centralized data center. Building on R2 lets K2 rely on object storage for durable state while its compute and storage capacity can scale separately, according to Cloudflare.
For developers, K2 adds another option to Cloudflare’s asynchronous services. It is positioned for high-volume data movement, retention and fan-out, rather than managing the lifecycle and retries of individual work items. The reported producer delay may matter to applications that prioritize low-latency writes over durable buffering.
serverless event streaming service
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How K2 Fits Cloudflare’s Services
Cloudflare says it first developed K2 as a durable buffer for Basin Pipelines. Pipelines uses a pull-based processing model, so events need to be stored before a processing engine reads, transforms and writes them to R2. Cloudflare says accepted Pipeline events must not be dropped, making durable storage important when processing is delayed.
The company distinguishes K2 from Cloudflare Queues, which are designed to track individual units of work and provide features such as retries, delays and dead-letter queues. K2 instead handles messages in batches, favoring high-scale movement and fan-out over message-level retry controls. Batching can improve processing efficiency but contributes to higher producer latency, Cloudflare says.
Cloudflare also describes Basin Pipelines as the better fit when the intended destination is object storage or Iceberg tables. It recommends K2 when developers want custom processing or need to write to other destinations. These product comparisons and recommendations come from Cloudflare’s own launch post.
““Today we are launching Cloudflare K2 in public beta.””
— Cloudflare
As an affiliate, we earn on qualifying purchases.
Open Questions on Performance and Access
The launch post does not provide detailed benchmark methods, workloads or comparisons behind the reported 99th-percentile producer latency. It also does not specify pricing, service limits, regional availability, or the precise retention options available in the public beta. Those details would affect how developers assess K2 for particular workloads.
Cloudflare describes K2 as durable and says its design relies on R2, but the announcement does not include independent testing or a full technical account of failure handling and recovery. The company says it plans to publish a technical deep dive later. K2 is in public beta, so its capabilities and performance may change as the service develops.
As an affiliate, we earn on qualifying purchases.
Technical Details and Beta Testing
Cloudflare says developers can create streams through its tooling and a getting-started guide linked from the launch post. The company also says it will share more about K2’s design in an upcoming technical deep dive. That material may provide further detail on the log architecture and how the service manages writes and reads.
For now, teams evaluating K2 can compare its batch-based stream model with Queues and Basin Pipelines, while checking whether the beta’s latency and service terms fit their needs. Cloudflare has not given a date for the technical post or identified a timeline for K2 to leave public beta.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is Cloudflare K2?
K2 is a serverless event-streaming service in public beta on Cloudflare’s Developer Platform. It stores events as an ordered log and lets consumers read them independently.
How is K2 different from Cloudflare Queues?
Cloudflare describes Queues as work-item oriented, with features such as retries, delays and dead-letter queues. K2 is designed for batch-based data movement, long-term retention and fan-out to multiple consumers.
What latency does Cloudflare report for K2?
Cloudflare reports about one second of producer latency at the 99th percentile in the initial release. The launch post does not specify the test workload or benchmark method.
What is K2 built on?
Cloudflare says K2 uses R2 object storage as the foundation for its partitioned, durable log, writing events in segment files rather than appending each event to an object.
Source: hn
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
