What I like most about ScyllaDB is that it’s :
Wrote in C++ and not Java
Very low latency (we’re talking microseconds not milliseconds)
Requires no extra tuning unlike Apache Cassandra
Faaaast due to the shard-per-core design which avoids locks, uses own cache, wrote in C++ (no JVM)
API is compatible with Cassandra
Capable of 2 million requests per second per node
I would not use ScyllaDB just yet for production workloads but there is an Enterprise Edition coming out soon, all of Cassandra’s features will be implemented by the end of Q2 2017 and I believe that the user community will grow rapidly.
Definitely one to watch closely!