# Metrics (/monitoring/metrics)



Unbind records how much your services use, with nothing to install. Every metrics page has the same four charts:

* **CPU**: usage over time.
* **Memory**: usage over time.
* **Network**: traffic in and out.
* **Volumes**: space used by [volumes](/services/volumes).

## Where to find them [#where-to-find-them]

* **One service**: the **Metrics** tab of the service.
* **A project**: the **Metrics** tab of the project shows the services of the current environment.
* **A team**: **Metrics** on the team page compares its projects.
* **The servers**: see [System Metrics](/operations/system-metrics).

## Reading them [#reading-them]

Use **Filter** to change the time range, from 5 minutes to 30 days. On project and team pages you can also pick which services or projects to show, and switch between **Separate**, one line each, and **Combined**, one line for the total.

A few patterns worth knowing:

* **Memory that climbs and then drops to zero** is a service hitting its [memory limit](/services/replicas-and-health#resource-limits) and restarting. Look for `OOMKilled` in the deploy logs.
* **Memory that only climbs** is usually a leak.
* **CPU that sits flat at its limit** means the service is being slowed down. Raise the limit or add replicas.

Metrics are kept for 7 days by default.