Skip to content

Fleetbench v2.1

Latest

Choose a tag to compare

@liyuying0000 liyuying0000 released this 20 Jan 17:21
· 3 commits to main since this release

Following the major architectural overhaul in v2.0, v2.1 improves the benchmarks in data fidelity, better observability, and framework stabilization.

Key Changes

  1. High-Fidelity Benchmark Data
    We re-implemented field sample logic. This ensures the distribution of message types, enums, and nesting levels more accurately matches the statistical profile of our production traffic. We also improved message type generation to better simulate cache pressure.

  2. Framework Stability & Parallel Execution
    The Multiprocessing Framework parallel controller now gracefully handles failed benchmark worker runs, preventing individual workload crashes from stalling the entire suite. It also now support passing additional flags directly to the underlying benchmark targets for more granular configuration of individual workloads.

  3. Observability & Result Processing
    Cache size overrides are now included in the output context. This helps verifying system topology when running on emulators or non-standard hardware.

  4. Bug fixes and dependency updates
    We fixed several issues in the Multiprocessing Framework and update third-party dependencies and CI environment configurations.

We hope these improvements help you get more accurate and reliable performance data. If you have any questions or feedback, please feel free to contact us. Happy benchmarking!