[svlug] 99.99% uptime...

[Alvin Oga]

hi ya hvrietsc

yup....people only care about data....
but need good hardware to do it....

ha is a good way as is clustering....but it will still go down
for whateeer software/deamon reasons for mroe than 15 minutes..
in a period of 8 years

and for disk access you need dual port from 2 different hosts
to get to the disk subsystem.... ( not from the same server )

"dynamic ffailover is sorta dangerous..cause whatever the
reason was that caused the failure might propagate to the
mirror/backup system tooo..... i rather find out why the puppy
failed before switching it over.....

- but data should always be written somewhere in the meantime...
  not necessarily the backup/mirror system

and yeh...this is fun stuff...to try to get to 99.9999% reliability...

c ya
alvin
http://www.linux-1u.net....1U raid5.... 500GByte of data in one 1U chassis...

> hvrietsc at yahoo.com wrote:
> 
> usually the server being available 99.99999 or whatever % is not the issue
> what the user really wants is his/her data and the services on them
> to be always available. to do this you need to a whole bunch of belts and suspender stuff:
> 
> dynamic multipathing to the data: two physical ways to get to your data in case one breaks
> HA clustering: more than one machine in case one dies with automatic failover of all relevant
> services
> clustering filesystem: so you can have the same filesystem available to all clustermembers
> journaling filesystem: so that if all else fails you do not need fsck
> replication of data: so if journaling fails you can get the data from someplace else
> 
> i know of a company that sells all this stuff and it works the same across solaris, hp-ux, nt
> and soon linux,but i forgot the name ;-)
> 
> 
> On Sun, Oct 15, 2000 at 06:44:33PM -0700, Alvin Oga wrote:
> > 
> > hi ya....
> > 
> > some folks seem to want to provide 99.9999% uptime...
> > ( new kids(company) on the block )
> > 
> > so i wanted to see what that means....
> > 
> > lets pretend that a server goes down for only 15 minutes
> > for whatever reason....swapp disks...add memory...
> > disk crash ..etc..etc...
> > 
> > I claim the marketing folks going around quoting 99.9999%
> > uptime is insane ????
> > 	- guess i'll pass on that company....
> > 
> > you'd have to keep that puppy up and running for 8 years and
> > still couldnt reach 99.9999% uptime...
> > non-stop except for 14 minutes.....
> > 	- how long does a reboot of the system takes ???
> > 	- how long does it take to e2fsck the 40Gb or 500Gb disks ???
> > 		- not much time for "maintenance"
> > 	- dont forget the kernel ticks tooo...( dies at 4xx days )
> > 	- and dont forget too that you have to drive to get to that dead box
> > 
> > -- adding more machines in the cluster for "HA" makes the problem
> >    even more complex ???? ( orders of magnitude easier to keep
> >    one server up and running and guaranteeing data saved onto 
> >    secondary backup hot-swap box that is also up and running
> >    	- change the ip# and your dead box is replaced ...
> > 
> > c ya
> > alvin
> > 
> > 
> > -------------------------------------------------------------------
> > assume....14 minutes of down time.... for the life of the server...
> > -------------------------------------------------------------------
> > #
> > # Daily...                      downtime...     uptime...
> > # --------
> > # 1,440 minutes per Day....     0.972222%       99.0278% uptime
> > #
> > #
> > # Weekly...
> > # ----------
> > # 10,080 minutes per Week...    0.138888%       99.8611% uptime
> > #
> > #
> > # Monthly...( 30 days )
> > # ----------
> > # 43,200 minutes per Month...   0.032407%       99.9676% uptime
> > #
> > #
> > # still reasonable... uptimes...
> > #
> > #
> > # Quarterly... ( 90 days )
> > # ------------
> > # 129,600 minutes per Quarter...0.010802%       99.9892% uptime
> > #
> > #
> > # Annually... ( 365 days )
> > # -----------
> > # 525,600 minutes per Quarter...0.002664%       99.9973% uptime
> > #
> > #
> > #1,051,200 minutes per 2yrs.....0.001332%       99.9987% uptime
> > #2,102,400 minutes per 4yrs.....0.000666%       99.9993% uptime 
> > #4,2048000 minutes per 8yrs.....0.000333%	99.9997% uptime
> > #
> > #
> > #......
> > 
> > _______________________________________________
> > svlug mailing list
> > svlug at lists.svlug.org
> > http://lists.svlug.org/mailman/listinfo/svlug
>