Infrastructure & Operations › Observability

USE Method

Utilization, saturation and errors for resources.

Backend Engineer · Senior Data Engineer · Senior

Also known as: USE method, utilization saturation errors, USE metrics

The USE method is a checklist for monitoring any resource — CPU, memory, disk, network interface, a thread pool. For each one, check three things:

Utilization — how busy it is, as a percentage of capacity: CPU at 70%, disk at 40%.
Saturation — how much work is queued waiting for it: run-queue length, I/O wait, a full connection pool. This is where trouble shows up first.
Errors — counts of failures: disk errors, dropped packets, failed allocations.

utilization:  70% busy
saturation:   queue growing        ← the early warning
errors:       non-zero            ← the hard failure

It complements the RED method, which is for request-driven services. Rule of thumb: USE for resources, RED for services. If you’re staring at a disk, a network link or a database connection pool, ask utilization, saturation and errors; if you’re staring at an API endpoint, ask rate, errors and duration.

The classic mistakes:

Watching utilization and ignoring saturation. A resource at 100% utilization but with an empty queue is fine; one at 60% with a growing queue is in trouble. Saturation is the leading indicator that something is about to fall over (see load average).
Applying the wrong method. USE on an HTTP service, or RED on a disk, gives you numbers that don’t answer the question. Match the method to the subject.
Only the obvious resource. The bottleneck is often not CPU: it’s a connection pool, a lock, a network link or a disk queue. Check each resource the request path depends on.
Ignoring errors. Non-zero error counters on a resource are easy to overlook in an average-based graph, yet they mean failures right now.

How to use it: when something is slow, walk the resources the request touches and ask the three questions for each, until you find the one that’s saturated. It pairs with per-process tools for a single machine and with capacity planning when you’re deciding whether to add more. Together, USE and RED cover resources and services — the two halves of a complete first look.