ZenMark Benchmark
A bounded, versioned benchmark for node capacity decisions.
ZenMark (Settings → System → Benchmark) measures what a node can actually do before you buy more of it — compute, storage and network are reported separately, because a machine that is fast at one is not necessarily fast at all three.
How it runs#
- The benchmark is bounded: it never pauses your workloads and it does not upload results to a leaderboard — the numbers stay in your instance.
- Scoring formulas are versioned (v1, v2), so a score from last year is comparable to one from today only when the formula matches; the run records which version produced it.
- Network quality is reported as its own dimension, kept apart from the compute/storage score — a great disk behind a saturated NIC told you nothing about throughput.
Using the results#
Run ZenMark on a new node before committing it to a fleet: the compute score says whether it can build, the storage score whether it should hold managed volumes, and the network score whether the agent transport and image pulls will keep up. Results and formulas export as JSON for your own records and spreadsheets.