Re: Got a different Linux box, it *still* keeps crashing

[z007400b@bcfreenet.seflin.lib.fl.us (Ralph Silverman)]

Kurt Hockenbury (khockenb@attila.stevens-tech.edu) wrote:
: Chris Osborn (fozztexx@nvc.cc.ca.us) wrote:
: : In article <Dpnr07.sK@nvc.cc.ca.us>,
: : Chris Osborn <fozztexx@nvc.cc.ca.us> wrote:
: : >I was able to talk a PC guy here into letting me borrow a P75 (a
: : >Gateway 2000 P5-75) with 16Mb RAM to use instead of the P100 that was
: : >crashing all the time.

: : It hasn't even been 6 hours, and the Gateway has crashed *again*. The
: : kernel is 1.2.13.

: : I've been getting a lot of people telling me that Linux is stable, and
: : the ISPs are using it exclusively, but I have *yet* to see a Linux box
: : run for more than a few days.

: Anytime I've seen stability problems, it's been hardware.
: It's plenty stable around here:

: Proxy web server and TIS firewall:
:  10:10pm  up 150 days,  8:14,  4 users,  load average: 0.08, 0.04, 0.06

: Web server (rebooted only because it was moved; up over 100 days before 
: that):
:  10:10pm  up 22 days,  6:36,  2 users,  load average: 0.00, 0.00, 0.00

: My desktop workstation:
:  10:10pm  up 100 days,  8:49,  9 users,  load average: 1.07, 1.08, 1.08

: : I find it hard to believe I've got a hardware problem when on two
: : seperate motherboards I'm getting crashes within a few hours to a few
: : days.

: : Is there no fix for this?

: Have you checked the external cache?  Try turning it off - sometimes 
: not-quite-fast enough cache ram can cause timing problems.

: 	-Kurt

--
************begin r.s. response*************


	linux is a clone of
		unix
	and shares its complexity...
	particularly regarding
		multiprocessing
	.

	the multiprocessing of unix is
	very demanding of system resources
	since
		process switching
	may ordinarily happen at intervals
		<= 1/10 second
	.

	stability of transfer to another process,
	and return to it,  are critical...

	the evolution of intel family cpu s
	was oriented,  into the early
		80486
	and earlier (286 386)
	to handle this demand...
	and largely these were very good for this...

	however...
	the actual history of the small computer
	did not directly lead to unix or other
	similarly
		multiprocessing os
	and standards in cpu and system design
	went to support other system forms...

	therefore...
	hardware reflecting these developments
	may not appropriately operate with 'unix like'
	environments...

	particularly...
		pipelining
	and
		cache schemes

	which may test well in
		essentially single processing
	environments
	may be effected...

	reputably...
	this problem is most virulent
	in cpu
	reflecting the generation next 'up'
	from the
		80486
	.


************end r.s. response***************
Ralph Silverman
z007400b@bcfreenet.seflin.lib.fl.us




--

Ralph Silverman
z007400b@bcfreenet.seflin.lib.fl.us