October 2015 Summaries

In a benchmark test conducted by the ScyllaDB team, ScyllaDB significantly outperformed Apache Cassandra across various workloads on a 3-node cluster with replication factor set to 3 and consistency level at QUORUM. The results showed that ScyllaDB achieved average operations per second of 1.93 million for write, 1.95 million for read, and 1.55 million for mixed workloads, compared to Apache Cassandra's 0.13 million, 0.05 million, and 0.07 million, respectively. This performance evaluation highlights ScyllaDB's superior capability in handling database operations efficiently compared to Apache Cassandra.

ScyllaDB achieved a new throughput record of 1.87 million transactions per second (tps) by increasing the number of loader virtual machines (VMs) during testing, which was conducted on the same hardware as a previous benchmark. This adjustment shifted the throughput saturation point, allowing the system to handle more transactions. As a result, ScyllaDB saw a 44% improvement in write operations, a 34% improvement in read operations, and a 16% improvement in mixed operations compared to previous results. These enhancements highlight ScyllaDB's growing efficiency and capacity in processing large volumes of transactions.