Work out nice priority effect instantly with clear inputs, formula shown and shareable results.
The Linux completely fair scheduler assigns each nice level a weight, where every step changes the weight by a factor of about 1.25. Ten nice levels of separation therefore gives roughly a nine-to-one CPU ratio, and twenty levels about eighty-to-one. Nice only matters when the CPU is contended; on an idle system it has no effect.
CFS weighting
weight = 1.25^(-nice); share = weight / sum of weights; each nice step changes CPU share by about 25 percent
No, it is proportional weighting among runnable tasks. For a hard guarantee use cgroup CPU quotas, which cap and reserve rather than merely prioritise.
Because it only affects CPU scheduling. Use ionice with the CFQ or BFQ I/O scheduler to prioritise disk access.