> 100 gigabyte hard disk is less than $200 while the last check on high > capacity tape drives turned up prices exceeding 4 times that for maybe a > quarter the capacity because advertised tape capacity is compressed > capacity.
I's an engineering tradeoff. But pricewise: Very true. Backups are not just made for hardware failures; i.e. an accidental rm -rf / wipes a RAID - no matter how well protected. Also; fragementation and other index losses are more of a problom than with thapes. > Worse, tapes don't last, they have a three year shelf life if > they are stored properly and the tape doesn't physically break when it > winds around the spools... You may want to check the web page at NASA aims on long term storage of remote sensing data. The situation is more complex. Also take note of what those remote sensing folks are doing - they typically create and processes gigabytes to terrabytes every so many days. Also - a tape can be repaired and you propably lost a few mg's or gb's. A dead disk, or an OS issue may cause you to loose the whole disk. > Is it possible to configure Amanda to backup to a harddisk or Raid > volume instead of a tape? What about CD-R, it's the cheapest media and 100 Gbyte is how may CD's :-) > My last search on tape drives suggested that a high capacity unit that > can handle 40 gigs per tape is between $800 and $1000. I could buy a Try e-bay; expect around 2500 dollars for a 1x24 tape changer; DLT or DAT. >From there your media scale in a known way. > lot of ATA hard drives for that kinda cash. ATA disks scale in funny ways; number of connectors, powersupply, supplier consistency, long term interface issues, low level OS depended file system issues. Quite a difference with sequential written data formats in a fairly independent manner. And when done with dump/tar/pax you can be reasonably sure that you can get your data back on any other odd system. Also you can take a tape off site; HD's go bad in multiple ways; tape goes bad in a few well understood ways, etc, etc... Lots of things going into this compromise. Add logging file systems and HFMS-es and the story gets even more complex. Again - there are some facinating problems described in papers of the ever cash-lacking scientific remote sensing communty - wher ethey have lots of data and want to make their little money go as far as possible. Dw.>
