[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: high load average



on Tue, Mar 06, 2001 at 11:21:07AM -0600, Dave Sherohman (esper@sherohman.org) wrote:
> On Tue, Mar 06, 2001 at 06:09:41PM +0100, Joris Lambrecht wrote:
> > isn't 2.00 more like 2% ? It is US notation where . is a decimal separator.
> > Not ?
> 
> You have the notation correct, but load average and CPU utilization are not
> directly related.  Load average is the average number of processes that are
> waiting on system resources over a certain time period; they could be waiting
> for CPU, for I/O, or for other resources.  (CPU does tend to be the biggest
> bottleneck, though, so a basic rule of thumb is that you usually don't want
> load to be much greater than the number of CPUs in the box.  

It *is* CPU.  These are processes in the run queue.  A process blocked
for I/O or another resource is blocked, not runnable (I think, I'm not
positive, but I'll bet my morning coffee on it -- which I *really* like,
and you'll want to give it to me anyway if I don't get it).

The significance of load average is that if you have more runnable
processes than CPUs, you have identified a system bottleneck:  it's now
possible to increase total system throughput by providing either more
and/or faster processors. 

Excessive swapping indicates the system is memory bound.  This isn't to
say that having a large amount of swapped memory is bad (it may or may
not be), but having a large number of processes swapping in and out of
memory is bad.

Not sure what the metric for I/O bound is.  Under Solaris, top would
report on I/O wait.  I could crack the O'Reilly system performance
tuning book and see what it says.

If none of the above are evident and things are still too slow, then
start optimizing your program(s).

> The machine I'm using starts killing off processes if load exceeds 6
> or 7; I wouldn't want to see it hit 100...)

It may not be all bad.  In certain cases, I believe Apache will spawn
large numbers of processes which manage to count against load average.
However, total system performance isn't actually negatively effected too
much.  I once took my UMP PII/180 box to a load of about 30 by running
multiple instances of computer v. computer gnuches....  That took a
while to clean up.

-- 
Karsten M. Self <kmself@ix.netcom.com>    http://kmself.home.netcom.com/
 What part of "Gestalt" don't you understand?       There is no K5 cabal
  http://gestalt-system.sourceforge.net/         http://www.kuro5hin.org

Attachment: pgpRTNv0Nd2Eq.pgp
Description: PGP signature


Reply to: