disadvantages of a linux cluster - 99.999

alvin at Maggie.Linux-Consulting.com alvin at Maggie.Linux-Consulting.com
Wed Nov 6 19:31:31 EST 2002


hi ya leif

yupp... marketing hipe ... that the "set of machines/cluster"
will provide 99.999% uptime ... not invididually

an individual machine going down for normal maintenance
or emergency failure is NOT counted in teh overal 5-nine's uptime

its practically impossible for one machine to provide
99.999% uptime...
	- my best machines been up over 800 days... non-stop
	but still went down (init 1 )for e2fsck once in a while
	and backup ....( uptime remains un-interrupted )

	- uptime can be faked :-)

	- uptime is interrupted by reset and power off
	and is oblibious to "init 1" to do maintanence

- so what is the definition of 99.999% uptime ??? 
	- marketing voo-doo :-)

	- if an indiviual PC goes down ( any reason ) it should
	be included in the 99.999 reliability/uptime
	( it costs time and $$$ to play with each PC )

c ya
alvin

#
# remember that a pc takes about a minute-or-two to power down
# and slightly longer to reboot -- lost couple minutes already
#
http://www.Linux-Consulting.com/HA/99.999.txt
	15minutes/day downtime ----> 99.0278% -- good enough for most --
	15minutes/week downtime ---> 99.8611%
	15minutes/month downtime --> 99.9676%
	15minutes/quarter downtime-> 99.9892%
	15minutes/year downtime ---> 99.9973%
	15minutes/4 years downtime-> 99.9993%  <<< ahh finally >>

	-- your machine cannot go down for more than 15 minutes
	over a 4 year period :-)

On 6 Nov 2002, Leif Nixon wrote:

> 
> So a single node failure doesn't count against the uptime, is that
> what you are saying? If that is the case, precisly what definition of
> "uptime" are you using?
> 

_______________________________________________
Beowulf mailing list, Beowulf at beowulf.org
To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf



More information about the Beowulf mailing list