[Subject Prev][Subject Next][Thread Prev][Thread Next][Subject Index][Thread Index]

How to find server performance

Hi Lokesh,

The load average (loadavg) is a measure of how many processes have
been ready to run but waiting for CPU availability in the past 1, 5
and 15 minutes.  It's a very rough estimate of system load: the higher
your loadavg, the more busy the CPU is.  However, it's neither
reliable nor translatable, since some machines come to their knees
with a loadavg of 1 or 2, while others I've seen keep chugging away
merrily with loadavg's of 50 or more.

Since system load isn't a single-dimensional parameter, you'd do best
to use the Linux process accounting features to measure load on
various components like CPU, RAM, swap, disk I/O, serial I/O, etc over
a period of time.  Then you can decide which are the bottlenecks (if
any) on your system and upgrade the appropriate component.

Have a look at the Process-Accounting mini-HOWTO (slightly outdated),
and the sysstat package (http://perso.wanadoo.fr/sebastien.godard/).


-- Raju

>>>>> "Lokesh" == Lokesh Bhog <lokeshbhog@xxxxxxxxx> writes:

    Lokesh> Dear all, Can anyone explain what is the meaning of load
    Lokesh> average in the top command and also how can one monitor a
    Lokesh> server for a week.

    Lokesh> How can it be determined that the server is loaded and the
    Lokesh> CPU needs to be upgraded?

    Lokesh> Regards, Lokesh

Raju Mathur          raju@xxxxxxxxxxxxx           http://kandalaya.org/