A framework and common set of workloads for evaluating and comparing the performance of different NoSQL and cloud-serving databases.
YCSB (Yahoo! Cloud Serving Benchmark) is an open-source framework for benchmarking the performance of NoSQL and cloud-serving databases. It provides a standardized set of workloads to simulate real-world database access patterns and measure key metrics like throughput and latency. The tool helps developers and architects compare different database systems objectively.
Database engineers, system architects, and developers evaluating NoSQL databases for performance-critical applications or conducting research on cloud-serving systems.
YCSB offers a vendor-neutral, extensible framework with scientifically rigorous measurement techniques, including detailed latency percentile analysis. It avoids common benchmarking pitfalls and provides reproducible results across different database systems.
Yahoo! Cloud Serving Benchmark
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Predefined workloads A-F model real-world access patterns like read-heavy and update-intensive operations, providing a consistent basis for comparison across databases.
Supports multiple database bindings through a common interface, allowing benchmarking of various NoSQL systems without modifying core code.
Emphasizes percentile analysis (P99, P99.9) and avoids averaging latencies, offering scientifically rigorous measurement for SLA compliance, as highlighted in the README.
Customizable workload properties and support for running multiple instances enable comprehensive and tailored performance testing, with detailed histogram outputs.
Requires manual database configuration for each binding, and building from source needs Maven 3, which can be time-consuming and error-prone for new users.
Outputs raw histogram data that must be processed externally, such as with HDRHistogram tools, for merging and extracting percentiles, adding overhead to result interpretation.
No graphical user interface or real-time monitoring, making it less accessible for teams preferring visual tools or automated, continuous feedback loops.