Tapes, Hard Drives, Power and Reliability
There is an argument developing in favour of a return to tape for large scale or long term data storage.
The falling prices of hard disks coming below tape per byte stored, coupled with speed advantages of disk have been pushing toward farms of hard drives for backup and retention of large data volumes.
The issues with backup to disk that are not well publicised are retention stability and power consumption.
A tape sitting in a slot in an autoloader consumes a lot less power than a hard disk, even one that is in power save mode in a MAID array. The power supply to the MAID array has to be able to meet the load of all disks seeking simultaneously, thus making it inefficient when the disks are idle or spun down.
The elephant in the room for disk backup is data retention stability. Single parity RAID 5 can be shown to be completely inadequate for data protection for current numbers and sizes of drives due to the non-recoverable read error rates (BER) failing to improve in line with drive size. Even with dual parity RAID 6 (RAID DP, RAID Z) long term storage of large volmes of data is subject to controller and filesystem errors (Sun ZFS has some fundamental advantages over SANs here). For sufficiently large data sets that are not regularly read (or scrubbed) the data on disk will also decay beyond the capability of the parity scheme to repair. This makes it necessary to store and maintain multiple copies of all data retained, increasing the cost and power.
Tapes show significantly better medium term stability than hard disk and when used in parity sets can achieve very high Mean Times To Data Loss.
The final advantage that tape still has is portability, where it is desired to store the data in geographically separate locations disk archive systems require significant and expensive bandwidth for data replication. To quote a sysadmin "Never underestimate the bandwidth of a van full of tapes".