Benchmarks Overview RandRW Benchmark Metadata-heavy Benchmark Multi-agent Scaling Reproducibility Benchmark Methodology

Benchmarks

Reproducibility

Every benchmark publishes its full contract.

Every BrewFS benchmark publishes its full contract:

Workload configuration (read/write mix, block size, concurrency)
Backend setup (data backend, metadata backend)
Cache state (cold / warm)
Mount options
Hardware and software versions
Commit hashes for BrewFS and comparison systems

Performance claims must be tied to specific workloads. If a number cannot be reproduced from its published configuration, it is treated as a bug.

Multi-agent Scaling

How BrewFS behaves with 1 to 16 concurrent agents.

Benchmark Methodology

How BrewFS measures and reports filesystem performance.