ScyllaDB: Extensive NoSQL Benchmarking Evaluation

In summer 2023, we worked with the database manufacturer ScyllaDB to conduct a detailed benchmarking study on MongoDB and ScyllaDB.

This involved more than just a theoretical technical analysis of the two products in terms of performance and scaling. We also carried out 133! performance measurements for these two database technologies.

ScyllaDB used the results of these studies for a large-scale marketing campaign.

About ScyllaDB

ScyllaDB Inc. is a US/Israeli database company that develops the database of the same name. ScyllaDB is based in California, USA. They have already raised over $100 million in external funding to build their large-scale database, which is based on Cassandra.

ScyllaDB started as an open-source project in 2014 and now also has a commercial version and a fully managed database-as-a-service service, ScyllaDB Cloud.

The Challenge

The scope of this benchmarking project for an internationally established database provider demonstrates our established standing in the database community, which values our measurement methodology, our insights and our neutrality.

In addition to the technical details and challenges, the major challenge for a benchmarking white paper is precisely this objective, fair and complete presentation of the results.

The aim was to measure a wide range of customer use cases, even in large scaling regions. We use our automated benchmarking framework for 133 measurements with data volumes of up to 10 TB.

The Solution

Our automated database benchmarking framework was able to carry out the diverse measurement scenarios, 5 workloads up to 3 scaling sizes and numerous technical excursions, such as long-term runs, data model or consistency measurements in the shortest possible time and at attractive conditions.

The Results

All the results of the two projects can be found on our website or as a PDF download on the ScyllaDB website.

In particular, the benchmark measurements with some technical "nuggets" are a highlight and represent a new level in database benchmarks.

The raw data of the measurements can also be found on GitHub.