Utilization
The share of a resource's capacity in use over a period: offered work divided by capacity. Queueing delay grows sharply as utilization nears 100%: in a simple M/M/1 queue the wait is proportional to ρ / (1 − ρ), so going from 50% to 90% busy multiplies the wait by 9.
Performance
In Quorum#
Each Node shows its utilization: offered work over the capacity of its tightest limit, whether CPU, slots, memory, connections, or network. Queueing delay grows as ρ / (1 − ρ), the Node warns at 70% and shows danger at 90%, and above 100% it drops the excess. The Bottleneck search weighs only Nodes above 60%, and an Autoscaler targets 40%, 60%, or 80%. The Saturation row on the Dashboard shows each Node as a gauge and as a 24-hour timeline, with the same 70% and 90% lines.