Load average: How high is 'high'?

You probably shouldn't compare load averages with other systems, as those numbers depend on the system (number of cores specifically). If you have 4 cores, 4.00 means CPU 100% utilization, lower means underutilized. Higher would mean some jobs are waiting to be queued. So find out how many cores you have and you can tell what a good load average is.

http://en.wikipedia.org/wiki/Load_%28computing%29

It will be a like-for-like comparison – four cores out of eight available on a dual Xeon box.

00:57:12 up 27 days, 3:29, 2 users, load average: 0.02, 0.05, 0.11

That's mine, while it was backing up too (aka heavy gzipping and gpging) backups are between half and 1 gig linode 512

Load average is merely a measurement of number of jobs in the run queue. On Linux machines these are running, runnable… and processes blocked on I/O.

A load average of 4 on a quad CPU system isn't necessarily high; it's not inherently much worse than a load average of 1 on a single CPU machine.

The part about disk I/O, however, is important. If your machine is swapping a lot (excessive I/O) or if the VPS host is overloaded on I/O then your load average may go up simply because it takes longer for each job to complete. I've occasionally seen this in the past on linode (but not recently).

But at the end of the day the load average is merely a number; what actually counts is site responsiveness and user experience. You could have a load average of 0.4 but if the site takes 20 seconds to return the home page then it's possible not performing to your needs. And that can depend on many more factors than load average.

@sirpengi:

If you have 4 cores, 4.00 means CPU 100% utilization
Careful, not quite. In Linux, it means 4.00 processes were on average running, waiting to be run, or in uninterruptible sleep (waiting on disk I/O for the most part). In other Unix systems, it means 4.00 processes were on average running, or waiting to be run.

Although it seems semantic, especially on Linux a 4.00 load might not be an indication of full CPU saturation. A lot of people fall into that trap of thinking, and it bites 'em. When a process is waiting on I/O, the CPU can go do other things, so on Linux (and in that Wikipedia) the number is a bit deceiving.

NFS lagging is a good example of how this number might get disconnected from CPU status.

jed is dead-on. Here is a good article that explains load average: ~~[http://www.lifeaftercoffee.com/2006/03/13/unix-load-averages-explained/" target="_blank">](http://www.lifeaftercoffee.com/2006/03/ … explained/">http://www.lifeaftercoffee.com/2006/03/13/unix-load-averages-explained/]( > In short it is the average sum of the number of processes waiting in the run-queue plus the number currently executing over 1, 5, and 15 minute time periods.