cachegrand-benchmarks

Welcome to the cachegrand-benchmarks repository! This repository contains the benchmark data for cachegrand, the world's fastest key-value store and cache! Here you'll find the raw benchmark data, usually generated with memtier_benchmark, that showcases the incredible performance of cachegrand as well as comparisons to competitor key-value stores like Redis, KeyDB, DragonflyDB, and more.

Repository Structure

• benchmarks: This directory contains the raw benchmark data folders.
• scripts: This directory contains scripts used for generating and processing benchmark data.

Benchmark Data

The benchmark data is organized in the following folder structure:

[platform_name]/[platform_version]/[benchmark_specs]/[node_number]/[thread_count]/[operation_type]

• platform_name, the name of the software being benchmarked, e.g., keydb, redis, cachegrand, or dragonfly.
• platform_version, the version of the software being benchmarked, e.g., v6.3.2 for KeyDB or v7.0.11 for Redis, this folder will also contain an additional README.md with the specs of the servers used.
• benchmark_specs, the test configuration used for the benchmark, e.g., d128-c5-pipeline64 or d128-c25:
  • d indicates the data size.
  • c indicates the clients per node, 25 clients per node with 2 nodes (n01 and n02) means that 50 clients have been used in the test.
  • pipeline it's optional and indicates the size of batches of commands.
• node_number, the "node number" folder (e.g., n01, n02, etc.) contains the benchmarks generated by a specific load generation server, the hardware in use will be the same for all of them.
• thread_count, the number of threads used in the benchmark, e.g., t1, t2, t4, t8, t16, t32, or t64.
• operation_type, the type of operation being benchmarked, either get or set.

For example, to find the benchmark results for cachegrand v0.2.1 using the c5-pipeline64 configuration, generated by the load server n01 with t1 thread for get operations, you would look in the following folder:

cachegrand/v0.2.1/c5-pipeline64/n01/t1/get

Benchmark Types

cachegrand's benchmarks can be of two types:

• Auto-generated, these benchmarks are conducted periodically to track cachegrand's performance over time and in general they will use the commit id of the version tested out, the folder name will start with "git-" to differentiate
• Ad-hoc, these benchmarks are generated specifically for a release, showcasing the performance improvements and features introduced in that release, and the folder name will match the version used (e.g. v0.2.1).

Latest Benchmarks Results

To make it easier to access the data contained in the repo, a number of charts have been put together and attached to the repository as picture.

Contributing

If you have benchmark data you'd like to contribute or if you find any issues with the existing data, please feel free to open an issue or submit a pull request.

License

The benchmarks data in this repository are licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License (CC BY-NC-ND 4.0) meanwhile the scripts might be licensed under different licenses (e.g. GPL3, BSD-3-clause, etc.).