• Test Reports
• ›
• Community Benchmark

Community Benchmark ​

Performance numbers, not capability checks. Each round measures real upload, download, list, stat and delete throughput across cloud providers, with statistical aggregates (p50, p95, stddev, min, max) over multiple runs.

The community benchmark is a separate workflow from the integration tests in this section: those check whether a feature works at all on a given backend, the benchmark measures how fast it works once it does. The two answer different questions and live side by side.

Submission flow ​

New reports go through the Community Benchmark Report issue template, one issue per report. The original call, axpdev-lab/aeroftp#177, is closed. The CLI ships aeroftp-cli benchmark since v3.7.3 (stabilized in v3.7.4) with strict sanitization: no hostnames, paths, credentials, bucket names, or account ids ever enter the JSON report.

Anyone can contribute their own region and connection type by running:

aeroftp-cli --profile "Your Profile" benchmark standard \
  --consent-publish --report bench.json

Then open a new issue with the template and paste the block printed between --- BEGIN BENCHMARK REPORT (paste into the GitHub Issue) --- and --- END BENCHMARK REPORT ---.

Benchmark behavior checked for v4.2.5 ​

4.2.6 did not publish a new community round. The notes below describe the 4.2.5 runner. They are not a new measurement.

The latency row measures metadata round trips, not upload time divided by payload size. It performs one untimed warm-up stat, followed by five timed stat calls. Compare the same operation, connection mode and payload when interpreting throughput.

The runner uses a scratch tree, and cleanup continues after cancellation with bounded cleanup steps. A cancelled write can finish late on some services, so the runner checks the scratch root again and reports cleanup it could not confirm. Trash entries are purged only when the run can identify its own items; providers that cannot safely purge them can leave recoverable trash. Inspect the report before removing anything manually.

Provider throttling and file limits still apply. Koofr native API and Koofr WebDAV are different benchmark cells, and Nextcloud concurrency is bounded by the server/session path. These implementation notes do not revalidate historical measurements below.

Rounds ​

Why the page exists ​

When AeroFTP claims that one protocol is faster than another on a given provider, that claim should be defensible. A single residential fiber line in a single timezone is not a credible base for a public protocol comparison page. Selection bias in the maintainer baseline is explicit and called out in each round's disclosure section. Real conclusions only come from aggregating across at least three regions and two connection types.