Mailing List
Home
Forum Home
Linux - General Red Hat Linux discussion list
Installation - Getting started with Red Hat Linux
Enterprise Linux 3 - Discussion of Red Hat Enterprise Linux 3 (Taroon)
Red Hat Linux 9 - Discussion of Red Hat Linux 9 (Shrike)
Red Hat Linux 7.3 - Discussion of Red Hat Linux 7.3 (Valhalla)
Red Hat Linux 7.2 - Discussion of Red Hat Linux 7.2 (Enigma)
Apache Web Server
Oracle database, Microsoft SQL server ...
Subjects
application/x mplayer2 plugin
RPM error: db4 error(16) from dbenv >remove: Device or resource
   busy
Command stream end of file while reading
X Windows problem (xauth)
Upgrading openoffice 1 1 rpm
FTP: connection refused
FTP: connection refused
mount: /dev/cdrom: is not a valid block device
Dell Precision 650, RedHat 9, no sound
how to trace the cause resulting in the crash of bind server
Virus on the list
UNINSTALL RPM MYSQL
usb pen drives: mounting as a user
broadcom network interface
make mrproper
sendmail configuration on redhat
Couldn 't open PID file /var/run/named/named pid Permission denied
Promise 378 controller
kernel 2 6 and /dev/sound/mixer not found
Problem using up2date
mrtg step by step howto/configuration for a newbie?
Compiling and Installing Kernel 2 6
Can 't locate module ppp0, can 't locate module ppp compress 21
HOW I CAN MAKE BOOTABLE FLOPPY DISKET
Lotus Notes under Wine
/etc/security/limits conf question
Intel E/1000 driver
Command stream end of file while reading
rpm database corrupt
qla2300 modules
 
Search:  
Power your search with and, or, +, -, or "some phrase" operators.
Meaning of 'load ' on Linux

Meaning of 'load ' on Linux

2005-06-16       - By Tom Sightler

 Back
Reply:     1     2     3     4     5     6     7     8     9     10     >>  

On Thu, 2005-06-16 at 11:42 +0200, Mimmus wrote:
> He said me that "load" on Linux (i.e. output of 'uptime' or 'top') has a
> different meaning than in other Unixes. Under HP-UX, I was comfortable with
> values under 2-2.5; he said that under Linux 10-12 is normal and that he saw
> working servers with load even of 1000 (!) using Oracle 'custom' kernels.

Well, I'm pretty confident that load under HP-UX and under Linux were at
least originally the same.  Basically, load the average number of
processes in the runqueue during the given intervals of time.  I'm
pretty sure both HP-UX and Linux use this same basic definition.

For example, our central database server currently has the following
load:

load average: 2.35, 2.29, 2.25

That means that for the last 1, 5, and 15 minutes that typically 2.25
processes have been runnable (or possible 'D' state) at any given time.
Since this is an 8-cpu system that's certainly no real concern as that's
roughly 25% of the total capacity.  You should be able to roughly bear
this out with sar.  With a load average of 2.25 I would expect to see
approximate 28% of CPU time used and sure enough, here the last two
lines of my sar output from my Oracle database server:

                 CPU     %user     %nice   %system   %iowait     %idle
10:10:00 AM       all     35.33      0.00      3.05      2.55     59.07
10:20:00 AM       all     21.33      0.00      3.59      2.72     72.36

Reasonably close to my prediction based from load average, but not
exactly the same because they don't quite cover the same time periods.

During very busy periods I see loads as high as 20-25 for short periods
of time (< a few minutes).

A load of 1000 sounds crazy, but may not be based on how it is
generated.  My guess is it was generated during some benchmark rather
than a production, but I suppose a highly concurrent environment with
1000's of concurrent users might generate that load.  I've run web
benchmarks that pushed the load average to 100+ during the benchmark on
a 2-way system even though response to the clients stayed acceptable.

Later,
Tom


--
Taroon-list mailing list
Taroon-list@(protected)
http://www.redhat.com/mailman/listinfo/taroon-list

Earn $52 per hosting referral at Lunarpages.