Understanding Linux Load Average: What the Numbers Really Mean

What is load average?

In Linux, the load metric measures the amount of work the CPU is handling, essentially the length of the process queue. Load average is the average of this metric over three time windows: 1 minute, 5 minutes, and 15 minutes.

You can view the current load average with commands such as w, uptime, or top.

The three numbers (e.g., "6.68, 7.23, 5.66") represent the average load over the last 1, 5, and 15 minutes respectively.

First value – 1‑minute average load.

Second value – 5‑minute average load.

Third value – 15‑minute average load.

Linux samples the load metric every 5 seconds.

Interpretation of load average values

Single‑core processor

Imagine a single‑lane road where each car represents a CPU task. When the road is mostly empty, load < 1 (idle). When the road is fully occupied, load = 1 (fully utilized). If tasks exceed the capacity, load > 1 (queueing).

Multi‑core processor

On a multi‑core server, a load greater than 1 does not necessarily indicate a problem because the load is distributed across multiple cores. For a 2‑core CPU, a load of 2 means both cores are fully utilized.

To see how many CPU cores are present, you can run:

grep 'model name' /proc/cpuinfo | wc -l

When to be concerned about load average

0.7 < load < 1 – healthy state; the system can still handle additional work.

load = 1 – the system is at full capacity; investigate any additional load.

load > 5 – severe congestion; tasks are queuing heavily and performance will degrade.

Which load value should you look at?

Start with the 15‑minute load. If it is high, then examine the 1‑minute and 5‑minute values to see whether the load is trending down.

If the 1‑minute load exceeds 1 but the 15‑minute load stays below 1, the spike is likely temporary. Persistent high values across all three windows warrant deeper investigation.

How to troubleshoot a high load

When you encounter a high load, consider checking:

Running processes consuming CPU ( top, ps).

I/O wait times ( iostat, vmstat).

Memory pressure and swapping.

Network traffic and bottlenecks.

Source: ITeye Blog – https://heipark.iteye.com/blog/1340384