Saturation
Saturation is how full a resource is; as utilization approaches 100%, queuing latency rises sharply.
Purpose
Identify resource bottlenecks and scale before saturation to avoid latency spikes.
Mechanism
USE method: Utilization, Saturation (e.g., queue length), Errors. Near 100% CPU, requests queue and latency spikes. Near disk IOPS limits, latency surges.
Use cases
- Performance analysis
- Capacity planning
- Bottleneck identification