SOCIAL: Backup option for PC's (and macs)

[Mike Kosim]

Hey gang,
I know some people asked about backup options for PC's some time ago.   
Recently, I came upon a backup service that lets you backup to a  
separate location, something that protects from both breakdowns of the  
computer and catastrophic events (like home fire).

http://www.downloadsquad.com/2008/12/03/backblaze-offers-truly-no-hassle-online-backups/


Like all backup services and devices, this assumes that the backup  
device is in a location that is safe, and that the backup equipment  
itself is reasonably safe.  So using quick-n-dirty probability, you go  
from 4% chance of losing data to a .02% chance of losing data just by  
using one method of backup.  Using an additional method (like backup  
to Time-machine *and* an offline backup like backblaze) drops to  
something like 1 instance of data loss in 1 million user setups.   
(that's pretty awesome reliability)

Curiously, condom failure rates are 12%.  (user and product failures  
combined)  go figure.

Calculation methods: (from a procrastinating grad student)

Either way, from my (very, very) amateur view of probability, say you  
have a 4% defect rate in a product (very high), and the backup is  
reasonably reliable (.5%).  Data loss here is defined as failure of  
all systems, so probability of failure for the whole system (with  
probabilities multiplied) becomes the probability for data loss for  
the total system.

Data loss in a population (of similar devices) goes from 4% (4 defects  
per 100 machines) to .04*.005= .0002 or 2 instances of data loss per  
10,000 machines.  Using an additional, similarly effective backup  
method is calculated (by me) as .0002*.005=.000001 or 1/1,000,000.  I  
think from a rough-calculation perspective, you can use this method  
because not all data is there to backup (pun intended) the reliability  
of each component in your backup system.  Additionally, I know there  
are inaccuracies because reliability of each machine is calculated as  
a mean-time-between failure.  That means systems like time-machine  
actually improve on murky hard drive MTBF numbers because they turn  
off the high failure components (hard drive) when not in use, reducing  
the duty cycle (time when failures can accumulate).

One more thing -- I think there is a lot of value to using multiple  
backup methods that are different in character.  That is, they shift  
the parameters of failure so that a single sweeping event is unlikely  
to effect the whole system.  That's why having one time machine backup  
is good... but having another time machine sitting on top of the other  
time machine is only marginally better . . . because coffee destroyeth  
both when spilled by a clutzy intern.

This is simplistic, and I welcome critical review & correction of my  
numbers.  I don't assume awesome backup reliability because I haven't  
really been able to find good numbers.  I assume awesome is six-sigma,  
something like 3.4 defects per million.