Here is some further information about my CD creation methods, which
will answer some of the queries, and a summary of the useful
information I received.
I used a two stage process to burn the CDs.
1. Create an Isoimage file under Irix on my data server using the
freeware tools: gcombust and associated components. The main reason
for starting this way is to create a CD which is compatible with
Macintosh (HFS), PC (Joilet extensions) and Unix (Rockridge
extensions).
2. Transfer the Isoimage file to a Mac G4 and burn the CD using
Toast. It is relatively quick to transfer a single large file and I
have a fast CD burner on the Mac. The Titanium version of Toast is
required to burn an externally produced Isoimage file.
After burning each CD I use the 'verify' function in Toast to check
the disk, and also navigate the file structure a bit to make sure it
looks good. I did not routinely extract data back to the NMR system
to check it fully.
Finally I used the Fellowes labelling kit and filed the disks in
old-style jewel cases.
I made two copies of each disk and gave one to the supervisor. In
some cases I made a third for the user.
When attempting to restore onto the NMR server I tried 4 different CD
drives - the SGI O2 internal drive, a Mac G4 internal DVD drive, a
Mac external CD writer, and a Dell PC DVD drive.
It looks as if the worst mistake I made was using the labels. This
was confirmed by several correspondents. I was also advised to use a
special pen to mark the disks designed for CD/DVD labelling, and not
a standard marker pen. My disks are all labelled.
Today I was asked to return some data to the server for a student who
is writing up. Unusually he had some data which was only on CD. My
disk had several files which were unreadable. The copy of the data I
had made for him was fine though, and was not labelled with a full
size label, though he had stuck some tape on and written on that...
So it looks as if the CD labels should not be used for the job they
are designed for.
The quality of disk may have also contributed, but I don't recall
what they have all been and they are now covered by labels. The first
disks I bought were from LaCie, from whom I also bought the burner.
The only other factor I can think of is that my first 16x LaCie CD
writer died and I replaced it with a 48x model. Perhaps the old drive
made some dodgy disks at the end of it's life. I don't think that all
the faulty disks match a narrow time period though.
So what have I learned?
1. CDs are not entirely to be trusted for long term storage. Someone
suggested that each disk should be tested in the drive which was
intended to read them. This isn't practical for an archive as the
disks should outlive the drives.
2. Don't use labels and only use the special pens.
3. Add extra redundancy by routinely making myself two copies of each disk.
4. I have no information that burning speed is a problem, but I will
in future burn disks at slower speeds, I've used 8x recently. Burning
is so quick that this doesn't really matter.
5. Use 'known name' brand disks.
6. Try and keep and archive of data on a 'live' file system. This
should ensure that the data is always available. If necessary make a
decision to retire data to CD (or perhaps DVD) only at a certain age.
The next question is how reliable are DVDs for this kind of storage?
A lot more data will potentially be at risk from a single bad disk.
Thanks for your input,
Phil.
--
Dr Phil Dennison
NMR Facility Director (949)824-6010 (office)
Department of Chemistry (949)824-5649 (lab)
University of California, Irvine (949)824-8571 (fax)
Irvine, CA 92697-2025 dennison_at_uci.edu
USA
Received on Fri Oct 29 2004 - 12:52:50 MST