Red Hat Linux load refering to?

dwarf007 · 02-13-2006, 08:44 PM

Does anyone knows what is "load" refering to when we issue the command "top" in linux?

Thank You.....

accessrichard · 02-13-2006, 09:09 PM

Good question. I learn something new everyday.

http://www.teamquest.com/resources/gunther/ldavg1.shtml

Apparently the 3 load averages are 1 minute, 5 minute, 15 minute averages of "active processes as a measure of CPU utilization" or something like that.

You mentioned the "load", however all I saw were "load averages", hopefully that is what you ment.

gilead · 02-13-2006, 09:11 PM

The load average (according to http://wiki.linuxquestions.org/wiki/Load_average) can be described as:

Quote:

load average is a measure used to tell how busy the CPU is. load average show the average number of processes waiting for execution. Usually there are three numbers presented where the first show average last minute, second average for last five minutes and the third show average load for the last fifteen minutes.

Including how busy the CPU is, as well as the number of processes waiting for execution seems to be important as there have been others here with high load averages but low CPU usage.

dwarf007 · 02-14-2006, 08:51 PM

Thanks for the info guys. 1 more question, Even we know how to read it but when will consider high load?

10, 9, 8
50, 49, 48
100, 99, 98

It will be good if I know this, I can do a capacity planning for the servers. What will be the max point for the servers?

Anyone knows, please help.

gilead · 02-14-2006, 09:12 PM

Sorry I'm no help here, but I'm not sure what figure is considered high load. I'm not usually near either of the 2 boxes that occasionally bog down here when they actually do it. On the one, sendmail logs a message when the load is too high to process messages and on the other, Oracle emails me when the CPU usage stays at 100%.

chrism01 · 02-14-2006, 10:34 PM

If you read the page mentioned by accessrichard (& the part 2), you'll see how it's defined, but ime (and I think I've seen this mentioned elsewhere), most admins are not happy if the nums are consistently over 2, especially for all 3 nums.
However, as pinted out, it really depends ... if you are getting 'fast enough' responses, then it's not a prob...
You'd prob also want to query the CPU percentages and basic mem info that comes with top and make you own rules as to what is a problem...
Ideallt Linux won't stop, just slow down as these nums increase ... this assumes you don't have a memory leak that fills swap space.
If you do write you own perf checker, consider using Perl & module Linux::Statistics.pm .