unix "top" load average numbers mean what? - Tilted Forum Project Discussion Community

giblfiz · 10-20-2003, 05:55 PM

on the first line of output from the program "top" in a unix top there is a stat "load average" followed by 3 numbers. (here is a sample from my box)

top - 18:49:18 up 2 days, 8:25, 13 users, load average: 1.53, 0.77, 0.53

I think that these numbers represent how many threads are waiting in the Que to be scheduled for the CPU, but I'm not even that sure about that. I'm pretty sure that in general its bad for any of them to be much above 1, and represents the system choking on to much work.

I would like to know exactly what each of these numbers represent. Its not in the man page, and I'm having a bit of trouble googleing for it because "top" isn't a very good search word.

any answers would be quite welcome. Its something that I have used for a long time with a vague intuitive knowledge for a while and never really thought about until recently. I asked a few of my more knowledgeable flesh-friends, and they didn't seen to know either.

devnull · 10-20-2003, 06:27 PM

*cough* read the man *cough* With that said...

uptime
This line displays the time the system has been up, and the three load averages for the system. The load averages are the average number of process ready to run during the last 1, 5 and 15 minutes. This line is just like the output of uptime(1). The uptime display may be toggled by the inter-active l command.

devnull · 10-20-2003, 06:31 PM

For future questions about that kind of stuff too, if there is a specific thing you are looking for in a man file but you can't find, it may be easier to do a search. You can do this by hitting '/' (the forward slash key) and typing what word or phrase you want to search for. In this case, I typed '/load' and hit enter and it highlighted the word load throughout the entire man file, making it very easy to find that specific part.

oberon · 10-21-2003, 06:22 AM

Also, I find that "load average" can vary from flavor to flavor (e.g. Solaris, Linux, FreeBSD) under similar conditions. This is, I believe, mostly due to differences in process scheduling and memory management.

Load average isn't really a great performance metric.

It is, however, relatively useful for determining if a system is overloaded and something needs an upgrade. But there are better ways to determine such things (vmstat, systat -vm, etc.). Unfortunately, such tools vary widely in availability on different flavors.

meff · 10-21-2003, 08:06 AM

Simply put, anything over '1' means things are queueing for processor time.. In top on Linux the higher processor eaters are on the top by default, so find it and fix/kill it..

meff · 10-21-2003, 08:08 AM

I forget to mention something else in top that may confuse you: memory usage. To get a semi-accurate image of how much memory you have free add the 'cached' amount to your mem free, and there you have it, Linux isn't really using that much memory for *applications*. Cached stuff drops immediately for application memory, but Linux is smart and uses unused memory for caching things to make them faster.

giblfiz · 10-21-2003, 04:11 PM

Thanks all you have been most helpfull.

Quote:

Originally posted by devnull
*cough* read the man *cough* With that said...

you know, it may be in the uptime manpage, but its not the top manpage. I would like to point out that the *first* thing I did was read the manpage.

10-20-2003, 05:55 PM	#1 (permalink)
giblfiz is a shoggoth Location: LA	unix "top" load average numbers mean what? on the first line of output from the program "top" in a unix top there is a stat "load average" followed by 3 numbers. (here is a sample from my box) top - 18:49:18 up 2 days, 8:25, 13 users, load average: 1.53, 0.77, 0.53 I think that these numbers represent how many threads are waiting in the Que to be scheduled for the CPU, but I'm not even that sure about that. I'm pretty sure that in general its bad for any of them to be much above 1, and represents the system choking on to much work. I would like to know exactly what each of these numbers represent. Its not in the man page, and I'm having a bit of trouble googleing for it because "top" isn't a very good search word. any answers would be quite welcome. Its something that I have used for a long time with a vague intuitive knowledge for a while and never really thought about until recently. I asked a few of my more knowledgeable flesh-friends, and they didn't seen to know either. __________________ Use the star one and you'll be fighting off the old ones with your bare hands -A Shoggoth on the Roof

10-21-2003, 06:22 AM	#4 (permalink)
oberon Quadrature Amplitude Modulator Location: Denver	Also, I find that "load average" can vary from flavor to flavor (e.g. Solaris, Linux, FreeBSD) under similar conditions. This is, I believe, mostly due to differences in process scheduling and memory management. Load average isn't really a great performance metric. It is, however, relatively useful for determining if a system is overloaded and something needs an upgrade. But there are better ways to determine such things (vmstat, systat -vm, etc.). Unfortunately, such tools vary widely in availability on different flavors. __________________ "There are finer fish in the sea than have ever been caught." -- Irish proverb

10-21-2003, 08:06 AM	#5 (permalink)
meff Vanishing, like I do.. Location: Austin, TX	Simply put, anything over '1' means things are queueing for processor time.. In top on Linux the higher processor eaters are on the top by default, so find it and fix/kill it.. __________________ Toy-like people make me boy-like.

10-21-2003, 08:08 AM	#6 (permalink)
meff Vanishing, like I do.. Location: Austin, TX	I forget to mention something else in top that may confuse you: memory usage. To get a semi-accurate image of how much memory you have free add the 'cached' amount to your mem free, and there you have it, Linux isn't really using that much memory for applications. Cached stuff drops immediately for application memory, but Linux is smart and uses unused memory for caching things to make them faster. __________________ Toy-like people make me boy-like.

10-20-2003, 06:27 PM	#2 (permalink)
devnull Crazy Location: Bit Bucket	cough read the man cough With that said... uptime This line displays the time the system has been up, and the three load averages for the system. The load averages are the average number of process ready to run during the last 1, 5 and 15 minutes. This line is just like the output of uptime(1). The uptime display may be toggled by the inter-active l command.

10-20-2003, 06:31 PM	#3 (permalink)
devnull Crazy Location: Bit Bucket	For future questions about that kind of stuff too, if there is a specific thing you are looking for in a man file but you can't find, it may be easier to do a search. You can do this by hitting '/' (the forward slash key) and typing what word or phrase you want to search for. In this case, I typed '/load' and hit enter and it highlighted the word load throughout the entire man file, making it very easy to find that specific part.