Load average

[Prashanth Waugh]

hi friends
few days before my sytem was showing load aveage near abt 22.00 but now a days it is showing only 1.00 to 2.00. But if check the processes by usin TOP or w|more command it is found that its %JCPU is more.how it is happening.is it possiblt to change some parameters so that system will show less load avrg even if the system averge is more.

[Andrew Merritt_2]

Hi Prasant,
Load average and %JCPU are different things.

The load average is a measure of how many processes are waiting to use the processor. See http://en.wikipedia.org/wiki/Load_(computing)#Unix-style_load_calculation for some background info on it.

JCPU is the percentage of CPU time used by the job.

Andrew

[Prashanth Waugh]

Thanks Andrew
I read regarding nproc variable which is a kernal parameter .so is there any relation between nproc variable and load average.

[Patrick Wallek]

No. The nproc kernel parameter is one of the "limit" parameters. nproc is the maximum upper limit for the number of processes that can be running on the system at any one time. If you have nproc number of processes running and try to start more, you will get an error.

As the number of processes increases, you **MAY** see your load average go up, but then again you may not.

[Prashanth Waugh]

Thanks Andrew and Patric you both clear my concept regarding load avg %JCPU and nprock variable .once again thnks

[skt_skt]

Did you had the real performamce problem when the load average was showing high?..If it varies you may have to identify what is going on the system whihc may cause the process to be in wait Q for CPU time.

[Michael Steele_2]

Got to go with Santhosh here. The classic definition of load average has always been the number of jobs in the run queue as reported in vmstat, sar, top and uptime. In fact, uptime is often embedded in .profile when logging onto root for the SA's convienience. As for the relationship between 'nproc' and maximum number of system and user processes in the kernel, I can see it but I also can't see a dynamically growing and reducing kernel parameter as a problem with load average. And typically I see a high load average associated to one or two hung or runaway processes that leads to an application restart. This is easy to check with the pcpu ps -o arguement in UNIX95 or prstat in solaris.