D1, Hyperdrive, Queues and Analytics Engine hit general availability
Cloudflare used the opening of its Developer Week to push three core data products to general availability: the D1 serverless SQL database, Hyperdrive for connecting existing databases to Workers, and Workers Analytics Engine, its time-series database. The company also announced new pull-based consumers for Queues and refreshed pricing for the analytics service.
The emphasis across these releases is making full-stack development viable on Cloudflare’s platform. A typical application stack — relational database for user data, object storage for content, a queue for background jobs, key-value storage for config, and time-series analytics for telemetry — can now run entirely on Workers, with data infrastructure operating across Cloudflare’s global network rather than in a single region.
D1 reaches production readiness
D1, Cloudflare’s global serverless SQL database, is now generally available. The GA release includes several long-requested capabilities:
- Database size limit raised to 10GB, with support for up to 50,000 databases per account
- New data export capabilities
- D1 Insights, an enhanced query debugging tool that surfaces which queries consume the most time, cost, or are inefficient
Pricing for D1 remains unchanged from the open beta announcement. The free tier stays in place, and the Workers Paid plan continues to include 25 billion row reads per month at no additional cost.
Cloudflare also previewed several features in development for D1, including global read replication that deploys replicas automatically to move data closer to users. The upcoming Replication API will handle replica scaling and consistency management, with session-based consistency available to ensure users see their own writes reflected while read traffic benefits from the performance gains of replication. Larger database sizes, expanded Time Travel branching capabilities, and APIs for dynamically creating and querying databases from within a Worker are also in the pipeline.
Hyperdrive goes GA, with free pricing on paid plans
Hyperdrive, which launched in open beta during last September’s Birthday Week, is now generally available. The service uses Cloudflare’s network to improve routes to centralized databases, keep connection pools warm, and cache frequently executed queries close to users. It works with popular database drivers and ORM libraries without requiring changes to application code.
Upcoming work on Hyperdrive includes support for MySQL, connecting to databases in private networks and cloud VPCs via Cloudflare Tunnel and Magic WAN, and more control over cache invalidation strategies for tuning between performance and data freshness.
Hyperdrive is now free for developers on the Workers Paid plan. That includes query caching, connection pooling, and the ability to create multiple Hyperdrives for splitting prod vs. staging environments or applying different cache configurations. Cloudflare said charging for Hyperdrive didn't make sense given the latency cost — six or more round-trips per request — of connecting to traditional database engines without it.
Queues adds pull-based consumers and message delays
Queues has added two features: pull-based consumers and message delivery controls. Any HTTP-speaking client can now pull messages from a queue via a new /pull endpoint to request a batch, then acknowledge them through the /ack endpoint. This lets queue consumers run anywhere, alongside existing legacy infrastructure. Cloudflare notes that internally, teams are using this to write device telemetry from its 310+ data centers and consume it from Kubernetes-based back-of-house systems. Messages remain in the queue until the consumer is ready to process them.
Queues also now supports delaying messages, both at the point of sending and when retrying failed messages. This allows tasks to be scheduled for the future and applies a backoff mechanism where upstream APIs have rate limits. Cloudflare says per-queue throughput will increase substantially in the coming months ahead of Queues reaching GA, with emphasis on reliability to prevent lost or dropped messages.
Workers Analytics Engine leaves beta
Workers Analytics Engine is now generally available, offering unlimited-cardinality analytics on Cloudflare’s infrastructure. It is built on the same ClickHouse-based system Cloudflare uses internally to observe service health, capture product usage for billing, and answer customer-specific usage questions — with at least one data point written on nearly every request to the network.
Pricing is based on two metrics:
- Data points written — each call to
writeDataPoint()in a Worker counts as one data point. All data points cost the same, with no penalty for adding dimensions or cardinality. - Read queries — each request to the SQL API counts as one read query. Query complexity doesn't affect the cost, and there's no need to reason about how many rows each query will read.
Free and paid Workers plans include allocations of both data points written and read queries, with set rates for additional usage. The pricing model is designed to let developers estimate cost by counting function calls and HTTP requests — "napkin math" rather than spreadsheet work.
The new pricing takes effect for everyone in the coming months. In the meantime, Workers Analytics Engine remains available at no cost, and workers can start writing data points with fewer than 10 lines of code.
