I also have been using MO disks for about 2 years with an optical jukebox.
The jukebox is a Pinnacle Alta 20GB jukebox that stores 16 1.3 GB disks(650
MB on either side), and is controlled by Pinnacle's Jukebox software that can
be accessed remotely from any workstation or from home. I also have been
using Pinnacle's PVFS (virtual file system software) software that is suppose
to provide transparent management of the optical jukebox and the file system
of the MO disks that have been setup in concatenation groups. However, I
found that the setup is not trival for SGI system compared to SUN systems
using Pinacle's PVFS software and installation scripts. Unfortunately,
Pinnacles's PVFS software is no longer being developed. They have replaced it
with 3rd party software (one software package is called Tracer from Tracer
Technologies, Inc.), but I am sure you can obtain their PVFS software if you
purchase a jukebox and really want it.
The PVFS software after it is setup is easy to use as long as you do not make
any mistakes, e.g. unload a MO disk from the jukebox that is still mounted.
If this happens you may have to rebuild the entire virtual file system for
that concatenation group, and that could take a few hours depending on how
many files are on the disks. You will also have to know which disk was
mounted so that you can fsck the disk.
I have set up our jukebox to contain 4 concatenation sets of 4x1.3 GB. This
way I can name each concatenated disk and then mount each one separately. In
my case, I then have 4 x 5.2GB concatenated disks. I then keep three of the
5.2 GB concatenated disks for multidimensional data and the last 5.2 GB
concatenated disk for 1D data. The only drawback with a virtual file system
is that it needs about 1K of magnetic disk space for each file on the
location map of the virtual file system. This can take up to possibly 200-300
MB of hard disk space if you have thousands of 1D files stored.
Overall I have found that this a good way for permanent archiving of all my
1D data and multidimensional data after doing data reduction as previously
described by many others recently. This also permits any clients. e.g Indy's
on our Bruker spectrometers to access the data remotely with XWINNMR.
Therefore, if you want to use acquisiton parameters and processing parameters
from some 2D data set, you can simply edc the file( as long as you are
running NFS and you have made the spectrometer host computer a client for the
PVFS server) to set up a new data set with the same parameters. The jukebox,
also provides 20 GB of disk space for larger data sets that can be easily
accessed and cp'd over quickly to another unix machine in a transparent
fashion to the user. This is clearly far superior to DAT tapes for retrieving
and archiving data.
The jukebox with MO disk drive ( which now can handle 4.3 GB MO disks or for
a 16 platter junkebox about 65 GB- wish I had that one) is approximately $6k
and another $2-4K for the virtual file system software ( Pinnacle's PVFS
software is only around $2K). The MO disks are about $80/1.3 GB disk.
Regards
Kurt
##############################################################
Kurt Wollenberg e-mail:kfw@lubrizol.com
Research Chemist ph#:(216)943-1200 ext2026
NMR Spectroscopy fax:(216)943-9022
Lubrizol Corporation
##############################################################