Redpanda validates a range of client libraries from the Kafka ecosystem, including libraries for Node.js, Python, Go, and more. Clients that have not been validated by Redpanda Data — but use the Kafka protocol — remain compatible with Redpanda Serverless, subject to certain limitations. Redpanda Serverless also provides in-console quickstart guides featuring franz-go, kafka-python, and KafkaJS.

New signups will get $100 in free credits, which expire in 14 days. To continue using Redpanda Serverless after the 14-day free trial period, users must enter a valid credit card within 7 days of trial expiration. Accounts remain active in perpetuity, but after this 7-day period, Redpanda will delete any data in an account that has not successfully registered a credit card.

Redpanda Serverless features usage-based pricing, billing users at the end of the month. We provide usage examples of small and large workloads in the pricing calculator above, but to summarize, we price Serverless usage on the following dimensions:

• Per GB in (write) 0.1 USD
• Per GB out (read) 0.1 USD
• Per GB retained 0.1 USD
• Per partition/hr 0.003 USD (no charge for replicas)

There is no SLA for the Limited Availability (LA) release of Redpanda Serverless. Other usage limits include, but are not limited to, the following:

• Ingress throughput of up to 10 MBps, with 0.5 MBps guaranteed
• Egress throughput of up to 30 MBps, with 1.5 MBps guaranteed
• Partitions: 100 max
• Consumer groups: 200 max
• Topics: 30 max

We are the most successful Kafka API implementation outside of Apache Kafka itself. What we learned while running the largest known Kafka workloads (5, 10, 20GB/s) is that in order to provide proper isolation, we needed to virtualize the entire TCP connection and session management. See the new architecture summary below.

We have been working on a full-on protocol virtualization layer for multiple years, so that it would require exactly zero changes to applications. Internally, Redpanda Serverless is a multi-tier architecture in the cloud — any cloud, any region upon demand.

In addition to broker-local, topic-aware, cluster balanced throttling mechanics, we needed a global fleet to adjust rates dynamically on a flat hierarchy of streams so we could absorb traffic spikes without needing to scale the fleet of backend cells. Redpanda Serverless’ mechanics enables the technology to do live TCP migration between cells within the Kafka-Metadata timeout duration allowing Redpanda to do live migration of terabytes of data in an instant. See the mechanics of whole-cluster restore for background information.

Redpanda has a 99.99% uptime guarantee SLA on our Enterprise tiers today. Please reach out to us if you’d like to upgrade. For Serverless, we currently offer a 99.5% uptime SLA; expect industry standard SLAs on General Availability.

Yes, it truly scales to zero. If you do not use Redpand Serverless, we do not charge you, period. In fact the initial bytes in the read-only “hello-world” topic don't even go against your free trial funds!

At launch, the Redpanda Serverless LA is a production-ready scale-to-zero Kafka API. Native integration and schema support are on the roadmap. Please integrate with one of our launch partners for a best-of-breed streaming experience.

We call this Read Replicas and have had this capability for over 2 years. The ability to read directly from S3 without incurring intercluster transfer costs is available today in our Enterprise tiers. See more on our docs.

Today, Redpanda Serverless does not provide managed connectors. Redpanda Dedicated and BYOC clusters do come with managed connectors. Please reach out if you’d like to upgrade.

Redpanda Dedicated and BYOC come with managed Redpanda Schema Registry (compatible with Confluent serializers/deserializers). Please reach out if you’d like to upgrade.

Serverless is currently only available on AWS in the us-east-1 (Virginia) and eu-central-1 (Frankfurt) regions. Ask for a region via #serverless in the community Slack team.