ScyllaDB as backend


I’m new to Grakn. I’m wondering if the platform supports ScyllaDB as a persistence backend? I believe Grakn is built on top of JanusGraph? If so, Janus supports ScyllaDB, as it is built on Cassandra.

Many thanks,


Hi @townsendb, at the moment there is no plan to support ScyllaDB unfortunately!


Thanks for the quick reply @ganesh. Is it something that could be considered in the future at some point? ScyllaDB has incredible performance and efficient resource utilization characteristics…and it is API compatible with Cassandra and targeted as a “drop-in replacement” for Cassandra.



We would be happy to exploit the performance gains of Scylla, however we looked at the support for Scylla at some time towards the end of last year and the conclusion at that time was that it would not be a simple drop-in replacement for Cassandra because of the status of Janus-Scylla integration at that time. If you have any more knowledge or insight on the topic or recent developments we would be happy to hear that.



Thanks very much for the details Kasper! I really don’t have any personal insights about the Janus-Casper integration. I simply read a few things online that suggested there was compatibility:
Scylla/Cassandra and Elasticsearch
As per compatibility matrix, the supported Cassandra version is 3.11 and the supported Elasticsearch version is 5. This repository uses Scylla instead of Cassandra, and according to the Scylla Cassandra Compatibility matrix we find that Scylla 2.0 is a drop-in replacement for Cassandra 2.1.8.

I’m at the stage of trying to select a graph database for a product. I spent some time in the DGraph community trying to understand the offering. I then moved on to JanusGraph and at that point I came across Grakn, which on first impressions looks very good!
Thanks for taking the time to respond. Additional performance gains are always appreciated, that’s why I wanted to get an early insight into Scylla integration.



What we definitely have planned is upgrading Cassandra to 3.X. Looking at the docs you linked, drop-in compatibility is suggested for 2.1.8 so it might be worthwhile reevaluating the adoption of Scylla. In general we are planning to focus on usability (reliability + performance) this year so replacing the backend might be one of considerations.

Hope that helps


Thanks Kasper. I’ll keep that in mind,