[thelist] how high constitutes troublesome load average.

Bob Meetin bobm at dottedi.biz
Sat Oct 10 09:54:20 CDT 2009

>> Keep in mind that there are lots of other factors that could make a  
>> machine
>> slow down (IO, DB, etc.), and load average is just one factor, but a
>> continually high load sure doesn't help.
> And while load average is a useful warning sign, it doesn't  
> necessarily follow that the user experience is of slow response times,  
> which is what actually matters.
> If you are seeing consistently high load averages, then that's  
> probably the point to do some evaluation of the real KPI.
> (mmm NFRs & NFT)
> Cheers
> Martin

Quite frankly the only concern I have is how the web server affects the 
viewers' experience.  When I am looking at 2 different servers, have 
access to view load averages on both, and tell them that server B 
"appears" to be delivering pages more lethargically than server "A" and 
of course server A's load average is substantially higher than B's, but 
then they seem to infer that it's below 20 now (and that is apparently 
in acceptable range) it leaves me a little breathless.  Deaf ears.

They don't accept my feedback as laboratory control.  The key, I hope, 
is in running an apples to apples comparison.  Yesterday I set up two 
identical installs of Joomla CMS on each server in subdirectories, one a 
swift template and one that I know is slow out of the box (4 total).  My 
thinking is that I will run some tests and record the results on each 
(YSlow) when the load average is low, then record similar results when high.

This is a similar test which I would like to do on a dedicated server 
(to eliminate other variables) but don't care to lease one for three 
months for a week's testing.


More information about the thelist mailing list