Fork me on GitHub


performance graph

The system scales up to weeks of usage, 100,000s of write operations and more than 10 GiB of managed data (a commit every 1-2s). We anticipate that the system scales further, but we support these performance numbers for all CRDTs with separate metadata like CDVCS, OR-Map and LWWR for now. We also realized 1000 write operations per second on a single peer for very small transactions if you properly batch them (e.g. every 10 ms).

We are interested in performance measures of real world applications to iron replikativ out as a solid default storage in front of business focused databases for fast domain specific queries. Ideally it will work similar to Apache Kafka or Apache Samza as a fairly care-free primary storage system building materialized views on streams of CRDTs, but in a decentralized and more decoupled fashion.