RS3 Storage
Store more with less overhead.
RS3 is a managed state store for Kafka Streams built directly on object storage with SlateDB. No disks to manage and no inter-AZ networking costs means it is 8x less expensive than traditional databases and simpler to operate, all while scaling elastically to infinity.
It's time to stop embedding RocksDB.
Prevent Downtime
Embedding RocksDB results in long restores, during which Kafka Streams cannot process events.
Improve Flexibility
Sizing an app with embedded state is difficult, and scaling is inelastic. So you either overprovision or run out of capacity.
Scale Globally
Checkpoints and multi-region object store access points unlock disaster recovery protocols previously unimaginable.
RS3: Responsive Streaming Storage Service
The case for a new storage solution.
Responsive’s significant contributions to SlateDB serve as the foundations for a new, cost-effective, highly-available storage service: RS3.
It allows your Kafka Streams apps to combine all the benefits of remote state with the elasticity and cost-effectiveness of object storage.
It allows your Kafka Streams apps to combine all the benefits of remote state with the elasticity and cost-effectiveness of object storage.
Infinite Scale
Using cloud-native object stores reduces replication costs to 0, reduces storage cost by up to 11x and makes managing RS3 a breeze.
Full Elasticity
RS3 servers are stateless which makes adding capacity instant. Scaling operations go from days to minutes.
Disaster Recovery
RS3 supports consistent checkpoints and timestamps, which serve as foundations for advanced functionality such as multi-zone disaster recovery.
BYOC Done Right
Since all durable state is in object storage, RS3 can be run very effectively in your network: no data needs to leave and the control operations on stateless services are simple.
Elegant Simplicity
Delegating durability to the object store and transactions to Kafka means the database layer is elegantly simple.
Can't wait to get RS3?
Apply for the RS3 beta program.