Benchmarking MongoDB vs ScyllaDB: Social Media Workload Deep Dive

BenchANT's comprehensive study evaluates the performance and scalability of MongoDB and ScyllaDB under a social media workload scenario, revealing that ScyllaDB substantially outperforms MongoDB in terms of throughput, latency, scalability, and cost-efficiency. Using a read-heavy workload modeled on YCSB Workload B, the tests demonstrate that ScyllaDB achieves significantly higher throughput and lower latencies than MongoDB across different workload distributions and scaling sizes, with ScyllaDB reaching up to 12 times higher throughput and significantly lower update latencies. ScyllaDB also shows near-linear scalability, with its throughput scalability closely aligning with theoretical expectations, while MongoDB experiences a larger gap between theoretical and actual scalability. In terms of cost-efficiency, ScyllaDB offers a superior throughput per cost ratio compared to MongoDB, especially under hotspot distribution conditions. Additionally, the study explores the impact of different data models and consistency levels on performance, showing that ScyllaDB benefits more from a key-value data model and weaker read consistency settings in terms of throughput increases.