Journaling file systems revisited

John Chambers jc at trillian.mit.edu
Fri Aug 9 11:14:34 EDT 2002


Chris writes:
| Another thing to consider... Backups are first and foremost meant to
| safeguard your important data.  Why would you backup to a media type
| (ext3/reiser in this case) that is less reliable than your current
| configuration??  I understand that it gives redundancy, but wouldn't you
| just feel better knowing your data was backed up on a proven fs?  An
| extreme example would be choosing floppies over DLT if price were no
| issue.  Go with the reliable media I say.

Of course, that is an argument that is widely used  against
tape  backup.   A  number  of  studies  have said that tape
backups fail more than half the time. The failure rate goes
up  radically  after  a decade.  Most of our civilization's
records from the 60's and 70's are lost because the  backup
medium  can't  be  read.   (And  video  archives from those
decades are decaying rapidly.)

This is the main thing that has led to  widespread  use  of
disk backup.  But there are problems there, because changes
in disk hardware have made most older disks  unreadable  as
the  drives  go  out  of  production  and the population of
working drives decreases.

I have a number of old backup tapes and  diskettes  that  I
can't read on any hardware that's available. Maybe some day
in the future some historian will stumble across them in an
attic  and  decide  to do what it takes to extract the info
from them, to learn about  what  obscure  computer  hackers
were doing way back in the 20th century.

There has been a significant move toward doing  backups  by
mirroring the primary data on disks located remotely on the
network.  This has the advantage that the data is available
quickly  without  specialized  software.  All you need is a
file server of some sort.

There has been a bit of a  move  to  use  CDs  for  backup,
though  there  are  problems  with 10-year-old CDs becoming
unreadable.

Linus Torvalds some years ago remarked that the best backup
method  is  to release your work to the public, and make it
so useful that others will back it up on their  disks.   He
had a point there.




More information about the Discuss mailing list