> Load average is more complex than number of (logical or otherwise) > CPUs vs the load average number. The reason being load takes into > account the processor state of "waiting for disk I/O".
Ah, yes. forgot about that. You can use a command like iostat to get more detailed info about I/O. The iowait field will give you the % time your CPU was idle due to waiting on system I/O (IE: reading from hard disk). > As far as my experience goes, when load is driven up, it is almost > always due to IO saturation, not CPU saturation. However, I don't have > much experience with PF systems, so they might have CPU saturation > issues. Interesting. My experience has been almost the opposite. But most of my workloads tend to be RAM centric and not disc centric which could account for that. Jake Sallee Godfather of Bandwidth System Engineer University of Mary Hardin-Baylor WWW.UMHB.EDU 900 College St. Belton, Texas 76513 Fone: 254-295-4658 Phax: 254-295-4221 ________________________________________ From: Matt Zagrabelny <mzagr...@d.umn.edu> Sent: Friday, September 9, 2016 3:07 PM To: packetfence-users@lists.sourceforge.net Subject: Re: [PacketFence-users] Server Load metric On Fri, Sep 9, 2016 at 2:37 PM, Sallee, Jake <jake.sal...@umhb.edu> wrote: > I always assumed that came from the same source that 'top' pulls from. > > > If I am correct then the number represents the workload of your system. In > simplified terms you want this number to always be less than the number of > processor cores in your system. > > > If you have a quad core system and you have a system load of 3.00 then you > are effectively running 3 of your cores at 100%. > > > If in a quad core system you have a value of 8.00 this means that you have > overloaded your system and there are 4 processes waiting while 4 other > processes are fully utilizing all the cores on your system. > > > Here is a bit more explanation if your interested. > > > http://www.howtogeek.com/194642/understanding-the-load-average-on-linux-and-other-unix-like-systems/ > > > TL;DR: the load score should always be less than the number of logical cores > in your system, if its not then your system is overworked and you need to do > something about it. Load average is more complex than number of (logical or otherwise) CPUs vs the load average number. The reason being load takes into account the processor state of "waiting for disk I/O". >From man proc: /proc/loadavg The first three fields in this file are load average figures giving the number of jobs in the run queue (state R) or waiting for disk I/O (state D) averaged over 1, 5, and 15 minutes. They are the same as the load average numbers given by uptime(1) and other programs. The fourth field consists of two numbers separated by a slash (/). The first of these is the number of currently runnable kernel scheduling entities (processes, threads). The value after the slash is the number of kernel schedulā ing entities that currently exist on the system. The fifth field is the PID of the process that was most recently created on the system. Thus, you could have a high load average and throw a bunch of CPUs at the issue and it doesn't change the problem one bit. It could be IO bound. As far as my experience goes, when load is driven up, it is almost always due to IO saturation, not CPU saturation. However, I don't have much experience with PF systems, so they might have CPU saturation issues. -m ------------------------------------------------------------------------------ _______________________________________________ PacketFence-users mailing list PacketFence-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/packetfence-users ------------------------------------------------------------------------------ _______________________________________________ PacketFence-users mailing list PacketFence-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/packetfence-users