lambda-bench

Benchmarks AWS Lambda cold and warm start performance across memory tiers and runtimes (Python vs Go). Produces percentile stats, cost-proxy calculations, and publication-ready plots.

[!NOTE] AI Transparency Disclosure: This project utilizes AI coding assistants to generate boilerplates, optimize benchmarks, and refine documentation. All critical logic and performance calculations are human-reviewed and verified.

Example output

Full example run → examples/go-vs-python/

Install

pip install lambda-bench

Or with uv:

uv tool install lambda-bench

Usage

Run a benchmark

lambda-bench run \
  --config functions.json \
  --cold-iters 15 \
  --warm-iters 15 \
  --region ca-west-1 \
  --output-dir results/

Re-plot from saved results (no re-invocation)

lambda-bench plot results/executions.json --output-dir results/plots/

Diff two result files

lambda-bench compare results/before.json results/after.json

Config format

functions.json is a list of function descriptors:

[
  { "label": "Python 512MB", "function_name": "my-func-python", "memory_mb": 512, "variant": "python" },
  { "label": "Go 512MB",     "function_name": "my-func-go",     "memory_mb": 512, "variant": "go"     }
]

IAM policy

The invoking principal needs:

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Effect": "Allow",
      "Action": [
        "lambda:InvokeFunction",
        "lambda:UpdateFunctionConfiguration",
        "lambda:GetFunctionConfiguration"
      ],
      "Resource": "arn:aws:lambda:*:*:function:*"
    }
  ]
}

Outputs

Development

uv sync --dev
uv run pytest
uv run ruff check src tests
uv run mypy src