Load average question.

[Silver_1]

Hi,

What is the meaning of these 3 load averages 1.10, 1.11, 1.08 ?


When we can say the machine is loaded?

[root@hp01 root]# w
18:09:53 up 91 days, 5:54, 1 user, load average: 1.10, 1.11, 1.08
USER TTY FROM LOGIN@ IDLE JCPU PCPU WHAT
root pts/0 10.10.10.10 6:09pm 0.00s 0.07s 0.01s w

Tx.

[Michael Steele_2 · Accepted solution]

average number of jobs in the run queue over the last 1, 5, and 15 minutes for the active processors.

As for your second question: "...machine is loaded...".

run and paste in the results of these commands:

sar -u (* cpu *)
sar -d 5 5 (* disk *)
sar -v 5 5 (* processes *)
sar -b 5 5 (* cache *)
swapinfo -tam (* memory swap *)
vmstat 2 20 (* virtual memory *)

[Ninad_1]

Ji,

Those are the figures for the load average on the system during the last 1, 5 and 15 minutes respectively. This gives you an understanding how the system was loaded over a period of time - because usually instantaneous values sometimes go very high and other times can be low.

The load average is similar to the CPU run-queue (which you can see using sar -q )
The load average/run queue - indicate average number of runnable processes on each CPU on your system at that time - Suppose you have 10 processes ready to run and if you have 4 CPUs then the load average/run-queue will be 2.5

Generally this figure if high indicates that there are more processes ready to run than your CPUs in the system can process at that moment.
But sometimes this figure in isolation can be deceptive - if there are a large number of short lived processes then this figure shows high value but usually does not affect the system performance.

Usually the term machine is loaded refers to CPU being loaded. So the total CPU utiisation , run-queue , GBL_PRI_QUEUE (In the OVPA/glance metric) are a good indicators to see if your system is loaded for CPU resource.
But you need to check for high memory, disk, network usages as well for a overall system performance.

Regards,
Ninad

[Chauhan Amit]

The system is not loaded in your case. Just adding to the above explanation.The 5 minute load average is, for a
uniprocessor system, the average number of processes in the run queue over
the last 5 minutes. It does not, generally, count processes that are
blocked waiting for resources other than the CPU however in some cases you
can get a large spike in load average from a NFS server going down that
the machine depends on.

For a multiprocessor system it gets a bit more complicated. HPUX uses the average of the average of all the CPUs (so a busy
system with an average of 4 processes waiting for each of 2 CPUs would,
under SunOS/Solaris have a load average of 8 while on a HPUX system have a
load average of 4).

High CPU load does not always mean that your CPU is at 100% all
the time. It is not to odd to see a load average of 2-3 with a average
CPU utlization of 50% or less because of lack of resources other then CPU
cycles.

There is a rule of thumb that the load average should not be greater
than the number of processors on the machine.

-Amit

[Vibhor Kumar Agarwal]

I'll go for 2.5X the no. of CPU's

[Silver_1]

Thanks for your inputs.