Load Average Normalized
Hello,
I have some doubts about the way that load is calculated in some UNIX flavours. Linux says that load average is not normalized, this means if I have a 2 CPU machine and a load of 1, the CPU is loaded 50% of the time. On the other hand, I think in HP-UX the load average is normalized. In AIX I couldn't get if is normalized or isn't. Could anyone give me a explanation if the load is normalized or not in each system (AIX, HP-UX, Linux, Tru64, Solaris)? Thanks. |
Never heard of normalized load averages. Are you sure HP-UX reports non standard values ?
In any case, Solaris reports a value that isn't affected by the number of CPUs. By the way, the load average isn't computed from CPU usage but from the runnable processes count at sample intervals. |
Quote:
Have a read of this - it has some errors, but will give a good idea of what the numbers actually represent. jlliagre's definition is for traditional *nix - Linux includes tasks in uninterruptible sleep as well. The calculations of the moving average is similar for all though. |
I've found on http://www.loudermilk.org/software/solaris-hpux.html that load average calculation in HP-UX is normalized:
Load average calculations are different. In HP-UX, unlike Solaris, Linux, or just about any other UNIX, load averages reported by uptime are divided by the number of processors in the system, so that a load average of 1.0 means the system has as many processes in the run queue as it has processors. That's all well and good, and it might even be a better approach than the other vendors, but once again HP-UX breaks with tradition and confuses administrators of other platforms. |
That's curious. HP-UX 11i v3 uptime manual page doesn't mention that normalization but is describing the usual Unix way (average number of jobs in the run queue).
http://h20000.www2.hp.com/bc/docs/su.../c02272185.pdf |
Quote:
|
All times are GMT -5. The time now is 06:35 AM. |