bento-bench

A utility for creating and running benchmarks against Bento clusters.

Installing

Binaries

See releases.

Build From Source

git clone github.com/zerokrab/bento-bench
cd bento-bench
cargo build --release

Running Benchmarks

A collection of prepared suites are available:

Source	Cycles	Count	Link
Order Generator (Tiny)	1M-10M	4	https://boundless-benchmarks.mintybasil.dev/suites/suite-og-4-1m-10m.tar.zst
Order Generator (Small)	100M-1B	5	https://boundless-benchmarks.mintybasil.dev/suites/suite-og-4-100m-1b.tar.zst
Order Generator (Medium)	1B	4	https://boundless-benchmarks.mintybasil.dev/suites/suite-og-4-1b.tar.zst
Order Generator (Large)	4B	4	https://boundless-benchmarks.mintybasil.dev/suites/suite-og-4-4b.tar.zst
Order Generator (Varied)	50M-4B	5	https://boundless-benchmarks.mintybasil.dev/suites/suite-og-4-50m-4b.tar.zst
Signal	50B	4	https://boundless-benchmarks.mintybasil.dev/suites/suite-signal-4.tar.zst
Kailua	12B-17B	4	https://boundless-benchmarks.mintybasil.dev/suites/suite-kailua-4.tar.zst

Note: Please open an issue if you would like to see other suites added.

To fetch and run a suite directly:

bento-bench run --fetch <link>

Example:

bento-bench run --fetch https://boundless-benchmarks.mintybasil.dev/suites/suite-og-4-1m-10m.tar.zst

Or download and extract manually:

curl <link> | tar -xv --zstd
bento-bench run --data-dir ./data

See bento-bench run --help for more configuration options.

Docker

docker run --mount <data-path>:/data ghcr.io/zerokrab/bento-bench:latest run --data /data

Timing Precision

bento-bench can time the execution of proofs by either:

• Checking TaskDB for start/end times (--check-taskdb)
• Recording wall clock time

For the best accuracy, TaskDB should checked. If it is not checked, --poll-interval may need to be specified to get accurate readings for smaller proofs.

Creating Benchmarks

Data Directory Layout

data/
├── manifest.json         # Benchmark index
├── images/{image_id}.elf # RISC0 ELF binaries
└── inputs/{input_id}.bin # Serialized input blobs

Manifest Structure

The data directory contains a manifest.json that describes each benchmark entry:

{
  "description": "My benchmark suite",
  "entries": [
    {
      "description": "A simple request (500M)",
      "image_id": "abc123...",
      "input_id": "def456...",
      "cycles": 500000000
    }
  ]
}

image_id and input_id correspond to filenames under data/images/ and data/inputs/. The cycles field is computed automatically during the prepare phase.

Note: On first run, if no manifest exists one will be created with an empty description.

Fetching Market Requests

bento-bench prepare-request \
    --request-id 0x1234...ABCD \
    --description "A simple request (500M)" \
    --data-dir ./data
    --rpc-url "http://node:8545"

This will fetch and save the image and input to the data dir, and append them to the manifest.

Import Local Images/Inputs

bento-bench prepare-local \
    --image /path/to/image \
    --input <input string> \ # Or --input-path to load from file
    --description "A local benchmark (1B)"
    --data-dir ./data

This will copy the provided image/input into the data dir, and append them to the manifest.

Uploading Suites

To tar and upload a data dir to an R2/S3 bucket, run:

R2_ACCESS_KEY=<key> \
R2_SECRET_KEY=*** \
./scripts/upload-suite.sh <data-dir> <suite-name>

License

This library is free software; you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation.

See the LICENSE file or https://www.gnu.org/licenses/lgpl-3.0.html for the full license text.