__/ [ Aragorn ] on Saturday 20 May 2006 07:19 \__
> On Saturday 20 May 2006 05:51, Ignoramus23298 stood up and spoke the
> following words to the masses in /comp.os.linux.misc...:/
>> Suppose, hypothetically, that I have a two CPU machine (or a dual
>> core machine).
>> Suppose further that there are two processes running, each an infinite
>> calculation loop (no swapping, I/O or network), each working 100%
>> Suppose that everything else is idle.
>> What would be the load average, 1 or 2?
> The number represented by the load average is only a representation of
> the average amount of processes that are waiting in the queue to be
> executed. It has nothing to do with how many CPU's there are.
Good explanation. There is no need to repeat it, but perhaps only to stress
that load average can go as high as you wish, depending on how estimates are
calculated over 30 (this depends on the distribution), 5, and 1 minute(s).
> On the other hand (and as /Handover/ /Phist/ has explained in his reply)
> an SMP machine - and that includes hyperthreaded CPU's and dual-core
> CPU's - is much better at multitasking/multithreading than any single
> CPU machine.
This also promotes applications that are designed differently, I suppose, as
to take advantage of multithreading.
> As an illustration, here's a pasted output from /top/ on my machine -
> it's a dual 32-bit Xeon with hyperthreading enabled, running as a
> workstation and a LAN server...:
> top - 08:13:10 up 3 days, 9:29, 1 user, load average: 0.24, 0.23, 0.15
> Tasks: 135 total, 2 running, 133 sleeping, 0 stopped, 0 zombie
> Cpu0 : 28.1% user, 0.7% system, 0.0% nice, 71.2% idle, 0.0% IO-wait
> Cpu1 : 2.0% user, 1.0% system, 0.0% nice, 97.0% idle, 0.0% IO-wait
> Cpu2 : 1.6% user, 0.3% system, 0.0% nice, 98.0% idle, 0.0% IO-wait
> Cpu3 : 5.0% user, 0.3% system, 0.0% nice, 94.7% idle, 0.0% IO-wait
> Mem: 4148260k total, 4126744k used, 21516k free, 26792k buffers
> Swap: 779072k total, 0k used, 779072k free, 3567968k cached
> The uptime is fairly low at the moment because I regularly need to
> reboot the machine due to a hardware failure. I have in the past
> already accomplished an uptime of 150 days with this machine - I was
> forced to take it down due to a power failure that outlasted my UPS
> reserve, if I recall correctly. ;-)
I've done 150 on a couple of machines. One killer was a colleague who decided
to shut down and, in the other case, it was the Kpilot daemon that refused
to synchronise with the handheld until a simple reboot.
> All functions well - i.e. the OS doesn't hang or anything - but the
> videocard regularly fills the screen with a solid color, both in the
> GUI and in character mode login consoles, and thus renders the machine
> unusable as a workstation. Quite an interesting thing if you're in the
> middle of typing an e-mail or a Usenet post... :-/
Yes, this happens to me whenever a screenshot is grabbed, once in 10 minutes.
I have become used to it, so I just carrying on working as if all is normal.
You could possibly 'nice -n19' the processes that you have control over. The
can make all the difference in the world.
> I keep a root console window open on my last virtual desktop and I have
> shortcut keys for switching from one virtual desktop to another, so I
> can easily and cleanly reboot the machine if need be.
When all else does not work, CTRL+ALT+F1 in Mandriva/KDE and off you go to
'top', 'kill', 'killall' and the various shortcut in 'top'. It has saved me
many times before. *smile*
> The videocard is due for replacement one of these days, though. ;-)
Passing on the kernel panic to somebody else on eBay? J/K
Roy S. Schestowitz
http://Schestowitz.com | GNU/Linux ¦ PGP-Key: 0x74572E8E
9:25am up 22 days 16:22, 11 users, load average: 0.17, 0.24, 0.49
http://iuron.com - next generation of search paradigms