Load calculation

[Prakash Nair]

Is their any formula to calculate the system critcal load?

[Antoanetta Naghiu]

There are tools available. top command seems to do a very good job. You can try glance (you have to pay for it but available for 60 days as a trial). vmstat (see man vmstat) and others (iostat, netstat, etc, see the other posts in the same issue).

[James R. Ferguson]

Prakash:

uptime gives the average number of tasks in the run queue over the last 1,5 and 15 minutes. This is a useful "quick" look.

...JRF...

[Alan Riggs]

All of the above plus sar.

And measurewear/perfView for a historic perspective (non-free)

[Brian M. Fisher]

sar (system activity reporter) can be setup through cron to capture a large range of performance data i.e. cpu, buffer, block device, tty activity, system calls, swapping, and many other statistics.

To set it up:
1) mkdir /var/adm/sa
2) chown adm:adm /var/adm/sa
3) create the following root cron entry:
0 6-18 * * * su adm -c "/usr/lib/sa/sa1 900 4"
This will run the collection progran every day between 06:00 & 18:00 every 900 seconds (15 minutes"

To view information use:
sar -? -f /var/adm/sa/sa??

where -? is sar option man sar
where sa?? is the date of the month i.e. sa01

Brian
<*(((>< er

[Anthony Goonetilleke]

Every five seconds, the kernel walks the proc table and counts how many runnable and imminently runnable processes there are at that specific time. The load average is an average of the last 12 totals.

The actual formula is statistical, so the kernel doesn't have to
remember each of the last 12 values. The formula is:

new average = [previous average * e^(-1/12)] + (new value * [1 - e^(-1/12)])

There is a more comprehensive explanation in Document Id W3564348

[Vincente Fernandes]

Try sar, top, glance(paid product), uptime, w,vmstat etc.