Method: how Vakna's numbers are measured
Note
This page is generated from a mock CLI, and its numbers are placeholders.
Verified against vakna 0.0.0-mock on
Every speed figure in these docs comes from the benchmark on this page, and every figure links here. The numbers are pending: this page holds the structure they will fill.
What does the benchmark measure?
Five timings, each from the moment the CLI sends the request to the moment the box answers:
| Measure | What starts the clock | What stops it |
|---|---|---|
| Cold boot | vakna create |
the box runs its first command |
| Resume | vakna resume |
the box's processes run again |
| Fork, each | vakna fork --count 50 |
each forked box runs its first command |
| Snapshot | vakna snapshot |
the snapshot can be forked from |
| Suspend | vakna suspend |
compute stops billing |
What hardware and region does it run on?
| Hardware | numbers pending |
| Region | numbers pending |
| Guest kernel | numbers pending |
| Client location | numbers pending |
How is it run?
Each measure runs 1,000 times a week, from a client in the same region, against boxes no one else uses. The page reports the median, the 99th percentile and the worst run, never an average alone.
What were this week's numbers?
| Measure | Median | p99 | Worst | Runs |
|---|---|---|---|---|
| Cold boot | numbers pending | numbers pending | numbers pending | numbers pending |
| Resume | numbers pending | numbers pending | numbers pending | numbers pending |
| Fork, each | numbers pending | numbers pending | numbers pending | numbers pending |
| Snapshot | numbers pending | numbers pending | numbers pending | numbers pending |
| Suspend | numbers pending | numbers pending | numbers pending | numbers pending |
Where is the code?
The harness, its raw results and the scripts that turn them into this table will live in a public repo. Its URL is pending.
What do the docs' examples show?
The examples on every page come from the mock CLI, and their numbers are mocked. Each one says so: the receipts print mocked.