Operations
These pages are about keeping SOF useful under real traffic, not just making examples compile.
If SOF is still new, start with:
Operating Principles
- start from safe defaults
- measure before changing queue or thread counts
- treat ingress choice as the first performance decision, not a late tuning detail
- treat outbound relay and repair traffic as an explicit posture decision
- prefer typed tuning profiles over copying environment bundles
- document host-specific changes so they are reproducible
Read This Section For
- choosing the right deployment mode
- understanding SOF's relay and repair behavior
- understanding what SOF can and cannot change about latency
- finding the safe baseline before expert-only tuning
- looking up one specific runtime knob quickly in the Knob Registry
The main split in this section is:
- Deployment Modes for mode and trust posture
- Tuning and Environment Controls for what to change only after measurement