Cassandra or ScyllaDB
What’s the best choice for storage, Cassandra or ScyllaDB , in terms of performance, scalability,availability…
4 Replies
depending cluster size, instance size ...
Okay thank you
Both of them are a great choice.
Cassandra may leverage some JIT compiler optimisations while ScyllaDB has higher socket read throughput as well as cache improvements and don’t have GC due to managing memory directly.
I didn’t compare Cassandra 4 to ScyllaDB 5, but folks usually tell that ScyllaDB has higher throughput and a lower latency.
Both can scale linearly to huge geo-distributed clusters, so it’s hard to compare their scalability between each other. I would say they are even more or less.
That said, they may have different features which you need to compare as well (for example ScyllaDB has shard-aware load balancing which usually brings better performance in highly loaded clusters).
If you are not an expert in these databases then both have Serverless providers available with different price points available for you. For ScyllaDB it’s Scylla Cloud and for Cassandra it’s AstraDB (these are the most popular ones while other providers may exist as well).
For availability it really depends on your consistency settings. If you consistency is ONE then your cluster is available until at least one node with a requested partition exists. If it’s QUORUM then the majority of the requested partition nodes should exist.
If you mean availability of cloud provided server-less deployments then it depends not only between providers but also between deployments using that provider. Usually varies between 99.86%-99.99%, but you should check with your storage provider for that.
Notice, if you deploy JanusGraph in the cloud alongside your storage backend - ensure you are using the same proximity, so that latency between JanusGraph to your storage backend is as small as possible because it will directly affect query performance.
This is a broad answer, but for more details and comparison between Cassandra and ScyllaDB it makes sense to use dedicated Cassandra/ScyllaDB communities which are specialising more on the storage details.
Thank you so much for your detailed response