Research¶
How well the kernel predicts measured serving performance, where it falls short and why, and how its laws are tested. For readers deciding how far to trust a prediction, or working on the model itself.
- Accuracy
- The two comparisons the kernel is scored by, the figures for this version, and what each comparison can and cannot show.
- How the laws are tested
- The kinds of test the repository relies on, and why most of them compare against the engine's own source or against an exact relation rather than against fixed numbers.
The places where the kernel knowingly prices something differently from the engine are listed on Known divergences.