0

I was looking at the system load of my shared hosting account.

$ uptime 
09:14:37 up 21 days,  5:38,  4 users,  load average: 3.85, 4.96, 5.21

I'm not worried by the first and second values, they represent peaks of work.

But 5.21 for the last 15 minutes seems a bit high. Or maybe not? How do I know that?

I'd like to read some answers to this last question.

UPDATE: then, what is the unit of measurement of those numbers?

Paolo
  • 16,955
  • 11
  • 31
  • 40
  • 1
    Did you give the man page a quick read? As for "is it a bit high?" it depends on your machine & the workload. If that's a 1024-core monster server, it's peanuts. – Mat Oct 20 '12 at 14:42
  • 1
    The man page is very clear. I stopped reading too early. – Paolo Oct 20 '12 at 14:48

1 Answers1

4

Load average is the number of processes who are in waiting queue for CPU time. These three values are the average number of processes in waiting queue for past 1, 5 and 15 minutes.

I would suggest you to read the following articles:

  1. Linux Troubleshooting, Part I: High Load
  2. Examining Load Average

I personally believes more on CPU IDLE Time than on load average values because your load average can increase if the number of processes are more but that doesn't mean that your CPU utilization is high, that just means more processes are in queue, it may be the case that some process is calling large number of system calls, which are waiting in queue.

The second link provides very good explanation. I really recommend you to read it once.

pradeepchhetri
  • 9,859
  • 12
  • 51
  • 59
  • In words of monitoring: If load/cpus >1 then warning; if load/cpus >2 then critical. So why is idle time "better"? – Nils Oct 21 '12 at 20:14