Understanding Linux Load Average: Meaning, CPU Usage, and Real‑World Examples

What Is Average Load?

Average load is the average number of processes in runnable or uninterruptible states over a period of time, not simply CPU utilization. A runnable process is actively using or waiting for the CPU (state R), while an uninterruptible process is waiting for I/O or other kernel‑level operations (state D).

In practice, the load value reflects the number of active processes; an ideal load equals the number of CPU cores. For example, a load of 2 on a 2‑CPU system means full utilization, while the same load on a 4‑CPU system indicates 50 % idle capacity.

Load Average vs. CPU Utilization

Load average includes both CPU‑bound and I/O‑bound processes, whereas CPU utilization measures only the time the CPU spends doing work. Consequently, a high load does not always mean high CPU usage:

CPU‑intensive workloads raise both load and CPU usage.

I/O‑intensive workloads raise load while CPU usage may stay low.

Many processes waiting for CPU raise both load and CPU usage.

Practical Scenarios

To observe these relationships, install sysstat, stress, and stress‑ng on a CentOS system (upgrade sysstat if needed).

Scenario 1 – CPU‑Intensive Load

Run a CPU‑bound stress test: $ stress --cpu 1 --timeout 600 In another terminal, monitor load with watch -d uptime.

Check CPU usage with mpstat -P ALL 5 20.

The load rises to about 1.6 while one CPU shows 100 % usage and negligible I/O wait, confirming that the load increase is driven by CPU consumption. Identify the offending process with pidstat -u 5 1, which reveals the stress process.

Scenario 2 – I/O‑Intensive Load

Generate I/O pressure:

# --hdd creates temporary files
# -i spawns workers that call sync()
$ stress-ng -i 4 --hdd 1 --timeout 600

Watch load with watch -d uptime.

Observe CPU stats with mpstat -P ALL 5 20.

The load climbs to about 1.7 while one CPU shows a high %iowait (≈63 %), indicating that I/O wait is inflating the load. pidstat shows the stress‑ng processes as the cause.

Scenario 3 – Over‑Subscribed CPU (Many Processes)

Start eight CPU‑bound workers on a 4‑CPU machine: $ stress -c 8 --timeout 600 Check load with uptime; it reaches >4, reflecting more runnable processes than CPUs.

Inspect per‑process statistics with pidstat -u 5 1. Several stress processes show %wait values around 50 %, confirming heavy contention and CPU overload.

These examples illustrate how load average captures overall system pressure, how it differs from raw CPU utilization, and how to diagnose the underlying causes using standard Linux tools.