On Fri, 10 Dec 2010 07:11:18 -1000, "John Keiser"
wrote:

I am in Hawaii where power surges are unfortunately common.

I live in the forest in the Santa Cruz Mountains. When the wind blows
or it rains, the branches and the power lines tend to meet, resulting
in power glitches.

If it makes you feel any better (probably not, but worth a try), after
every storm, I get a few calls from customers with hung network and
entertainment equipement. DSL modems, cable modems, routers,
switches, IP phones, wireless, computahs, security systems, DVR's,
printers, TIVO, etc. Just about anything with a microprocessor inside
can be made to hang. D Yuniskis covered races and hazzards so I won't
go there. Add to that the joy of memory (RAM) glitches. When the
power fluctuates, one of the most sensitive components is the common
serial or dynamic RAM commonly found in almost everything. A
momentary magnetic pulse from a nearby power xformer is usually
sufficient to produce a large enough field to flip a few bits. You
may not even notice that a few bits have been flipped until perhaps
days after the power glitch, when the operating system decides to use
those memory cells, and finds them in a bizarre state. This is why
many servers have ECC (error correcting) RAM.

The problem of unpredictable processor operation is well known as are
some of the band-aids. For low end hardware, usually nothing is done.
Just power cycle the box if it hangs. Some clever programmers add in
a watchdog timer, which monitors the state of some manner of commonly
updated register (i.e. the RTC) and reboots the device if it goes
comatose. While clever, it's not very reliable as the dead-mans timer
is part of the same processor that it's trying to protect. An
external watchdog timer works much better. It usually receives a 1
PPS (one pulse per sec) signal from the processor. If that
disappears, it's reboot time.

It isn't just power line glitches that cause hangs. Cosmic rays,
alpha particles from radioactive components, external fields, and bit
rot all contribute to the general lack of uptime.

Anyway, try not to worry too much. Features and functions are added
faster than bugs get fixed, so reliability and uptime rapidly some
minimum acceptable value. This value is usually set by when the
support phone starts ringing. When the customer complaints arrive,
it's probably time to fix the problem. Otherwise, few people complain
about ocassional hangs, crashes, and reboots.

--
# Jeff Liebermann 150 Felker St #D Santa Cruz CA 95060
# 831-336-2558
# http://802.11junk.com
# http://www.LearnByDestroying.com AE6KS