Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 13 Dec 2001 09:38:48 -0800
From:      "Bruce A. Mah" <bmah@FreeBSD.ORG>
To:        Holtor <holtor@yahoo.com>
Cc:        stable@FreeBSD.ORG
Subject:   Re: Load Averages 
Message-ID:  <200112131738.fBDHcm235905@bmah.dyndns.org>
In-Reply-To: <20011213172108.10411.qmail@web11606.mail.yahoo.com> 
References:  <20011213172108.10411.qmail@web11606.mail.yahoo.com>

next in thread | previous in thread | raw e-mail | index | archive | help
--==_Exmh_-1977408144P
Content-Type: text/plain; charset=us-ascii

If memory serves me right, Holtor wrote:
> I still fail to see why my systems loads went from
> 1.50 - 2.00. There's over 250 processes constantly
> running in "select" state. Loads are now almost always
> 0.00 and sometimes touching 0.10

If a process is in "select" state, it shouldn't be contributing to the
load average...it's blocked waiting for something to happen.  The load
average is the number of processes in the run queue (meaning the number
of processes ready to run), averaged over some time interval.  The w(1)
manpage will give you a few more details about this.

With the old load average computation, it was possible for the sampling 
of the queue length to become synchronized with processes that ran 
periodically.  This would make the load average look higher than it was 
(because the only time that the queue length got sampled was when these 
periodic processes were woken up).  The random jitter is an attempt to 
avoid this synchronization, thus estimating the run queue length more 
accurately.

> I'm confused as to what problem this solved besides
> creating problems.. Now I have no idea what the real
> system load is. Surely 0.00 load is not proper for a
> system running so many things.

I'm sorry, but I have the feeling that you don't quite understand what
the load average is.  Saying that "my machine has 250 processes, it
can't have a zero load average" doesn't mean anything unless you know
what those processes are doing and how they're using the system
resources.

> Reason being I use load averages to determine if a
> computer needs upgrading. When things go above a
> constant 2.00 or higher it either means upgrade to
> better hardware or reduce the amount of things running
> on that server to another server.

The load average by itself isn't really a good metric of system
performance.  For example, it won't help you find disk or network
bottlenecks.  If it was me, I'd take a look at developing some
performance metrics that depended on specific applications I was
running, and some values that were (or were not) acceptable to me. (For
example, how many Web hits can the machine process per unit time, for
some offered workload?)  I'd also use some of the other system metrics
(such as CPU utilization) for additional corroborating evidence.

> Maybe i'm missing something.. but this seems to only
> have broken things not fixed anything.. 

I think you're missing something.  :-)

Bruce.




--==_Exmh_-1977408144P
Content-Type: application/pgp-signature

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (FreeBSD)
Comment: Exmh version 2.3.1+ 05/14/2001

iD8DBQE8GOeo2MoxcVugUsMRAgCnAKDUgEv7L0gHQ1SidGOzNCMrIU46ogCfSreZ
tEro3w+k7tEEqxhALzwbBFE=
=0dkj
-----END PGP SIGNATURE-----

--==_Exmh_-1977408144P--

To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-stable" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200112131738.fBDHcm235905>