Re: safe load average
On Wed, May 01, 2002 at 10:15:36PM -0400, Seneca Cunningham wrote:
> After the sound of my system thrashing being a better alarmclock than
> the conventional variety, I started checking my system's load average
> (among other things) periodically. From what I saw in uptime, my average
> load average tends to range from 1.0 to 3.0, with occasional spikes up
> to 6.75 (which I doubt is healthy for this P100), and a minimum of about
> 0.5.
>
> I know that uptime(1) says that "Load averages that are consistently
> high (that is, all three number reported by uptime are higher than 1)
> indicate an overtaxed system." (I just noticed a typo in the manpage).
> And I know that 1 is the "maximum processor performance on a single
> processor system.".
>
> The thing is, is that this system is a single processor system, and I
> think it has a slight temperature problem. As it is, this system is a
> laptop with the heatsink directly below the keyboard (so using ice or
> or adding fans would be somewhat inconvenient), and by how hot my
> PCMCIA cards are when I remove them, I half expect "CPU0 on fire" to
> come up. The maximum operating temperature listed in my manual is 35C
> (internal or external, I'm not sure which, but I'm hoping external),
> while the cards feel like they're at least 45.
>
> So, something I was wondering about would generally be considered a
> maximum "safe" load average. I have had some problems with some hardware
> that look almost as if this machine wants to become a toaster (hopefully
> the hardware wasn't damaged, only confused, by the heat).
I don't think it really matters. There is no difference in actual
CPU work after the loadavg exceeds the number of CPUs in the system.
That is, at load 1 the CPU is working 100% of the time[1] and at
load 100 the CPU is still running 100% of the time. It isn't as if
the CPU is somehow exceeding its rated speed!
I'm sure the 35C temperature specification means the maximum ambient
temperature under which the system was designed to function.
Junction temperatures inside the integrated circuits are usually
specified up to and including 95C. Everything should be peachy even
if the internals are at 50C, although this will somewhat reduce the
operating life of your hard drive.
-jwb
[1] A load of 1.0 actually indicates that every time the scheduler
was invoked, there was 1 process runnable. It doesn't mean that the
CPU ran in a tight loop for the entire time.
--
To UNSUBSCRIBE, email to debian-user-request@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Reply to: