I've frequently witnessed load average used to argue a server needs faster CPU's. While load average is an interesting number, it is not the end all measure of a servers performance.
Load average is a moving approximation of the number of runnable processes over some period of time. The UNIX command top typically shows three values corresponding to one, five, and fifteen minutes. Here is an example from my desktop system:
It is important to note that a runnable process does not imply it is running. Processes waiting on network or disk IO are runnable. I bet processes playing memory games with themselves, like Java threads, are also runnable even though they are not running. Runnable is a state recognized by the operating system job scheduler. Running is another recognized state.
The moral of the story here is that a high load average is not always a problem. A bottleneck may exist or it may not. Some administrators and users fixate on load average and CPU utilization. Just remember that a load average is only one measure of one part of what makes a computer tick.
load averages: 0.16, 0.13, 0.09Image via Wikipedia
The moral of the story here is that a high load average is not always a problem. A bottleneck may exist or it may not. Some administrators and users fixate on load average and CPU utilization. Just remember that a load average is only one measure of one part of what makes a computer tick.
![Reblog this post [with Zemanta]](http://img.zemanta.com/reblog_e.png?x-id=53268baf-bdac-4a02-8f6b-20eca2005e78)
Leave a comment