Work out load average interpretation instantly with clear inputs, formula shown and shareable results.
Linux load average counts processes both running and in uninterruptible sleep, so it mixes CPU demand with disk I/O wait — a load of 12 on 8 cores may mean CPU saturation or a slow disk. Dividing by core count normalises it, and separating the I/O wait contribution is what tells you which resource to fix.
Load normalisation
load per core = load average / cores; I/O contribution ~ load x I/O wait share; CPU contribution = load - I/O contribution
Not necessarily. A batch system running at 1.5 per core is efficiently utilised, while an interactive service at 1.2 per core already has visible latency.
Historical definition. It makes load a measure of total system demand rather than CPU demand, which is why pressure stall information was added as a clearer signal.