From: Brewer, Edward (BREWERE@OD.NIH.GOV)
Date: Tue Jan 28 2003 - 12:00:37 EST
Admins
Here is a conversation thread between Charles Ballowe and myself:
Response from Charles Ballowe:
I think the RUNQ is a current snapshot of the queue length, while the
average
is a sliding average. I think the math you are using could be slightly wrong
as well, but my statistics skills are slightly rusty. I don't remeber how
to properly handle sliding averages to get a better average.
Curiousity tells me, run the script for 60 seconds with a 1 second
interval and see if the average of RUNQ is roughly close to the last
AVG60?
Unfortunately I don't have a better answer for you.
-charlie
From: Lee Brewer
Thanks for your response. I think the math may hold (but I will check)
Here is a sample of the output
RunQ 5 30 60
0 2.4 2.41 2.39
0 2.73 2.57 2.49
0 2.39 2.46 2.45
0 2.82 2.61 2.54
0 2.55 2.6 2.57
0 2.42 2.41 2.46
0 1.95 2.13 2.27
2 4.61 2.37 2.29
1 1.3 1.73 1.93
0 1.22 1.44 1.67
This is from collect with an interval of 30 seconds....so this represents 5
minutes.
That means that the run queue value should be the average of the run queue
over 30 seconds (I think that is what collect is doing). Now the values for
5/30/60 are an average of the average values for the last 5/30/60 over the
30 second interval.
As you can see the run queue is close to 0 during this 5 minutes...
What do you think?
>From Charles:
Oh, yeah -- RUNQ is just the queue of Runnable processes that aren't
currently running. Load average is the average number of processes that
are runnable, including those that are currently occupying the processor.
To all:
I think that answers the question, but I will keep it open for any other
responses.
Lee Brewer
This archive was generated by hypermail 2.1.7 : Sat Apr 12 2008 - 10:49:05 EDT