What load average means UNIX/Linux systems

Kategorie: Linux
Von: Marco Engeler

[engeler@debian ~]$ uptime 
 09:43:14 up 16 days, 22:41,  1 user,  load average: 2.53, 1.84, 0.98

The UNIX commands w, top and uptime print out the load average. But what do those three numbers mean?

Unix systems count the number of processes running on or waiting for the CPU at regular intervals (every 5 seconds on Linux. Note: Linux also counts processes in the uninterruptable sleep state, e.g. a process waiting to access a disk). The numbers gathered are then averaged over periods of 1, 5 and 15 minutes. The average is not the well known arithmetic mean but an exponentially damped weighted average.

The value of 2.53 in the above example means that during the last minute there where on average 2.53 processes running. Which means there were more processes than the CPU can handle at once and 1.53 process had to wait on average for execution.

Simply put the higher above the number of processors the load rises the more the systems is under stress. (Note on multicore systems you should take the cores into account as individual processes.)

How significant a high load is depends on the circumstances. On a server where many maintenance jobs run at night time when no users are connected a high load may be no problem.

---