There Is A Place For Tape

Some in the storage biz, including my compatriot George, believe optical disks and tape have seen their day and are no longer appropriate for storing archive data. They contend that magnetic disk based solutions are the appropriate repositories for not just the active, but also the deep archive.

If they're right, there's no place in the modern data center for these venerable media types. I've already made the case that we make backups to restore but we make archives so we can retrieve that data later, sometimes much later. If tape isn't appropriate for the long retention archive, why use it for backups which should have a much shorter time horizon.

In part they're using statistics from market researchers, who report that 30%, 40% or even 60% of attempts to restore data from tape fail, to argue that tape is unreliable. That argument of course ignores the nature of market research, where the Gartners, Forresters and IDCs of the world generate these statistics by having users answer surveys.

First of all failures -- because they're painful and cause us as system admins to restore from tertiary repositories, or reconstruct data or worst of all tell the business units their data is gone forever -- are more memorable than successful restores. Admins answering a survey question from memory will overestimate how frequently their attempts fail.

The other factor that those who assume a high restore from tape failure rate equals low tape reliability tend to ignore is that a restore failure isn't necessarily a tape read failure. In fact, most restore failures I've seen since we started using DLT and LTO technologies have been either failure to identify or locate the right tape or attempts to restore data from applications that were no longer in use, like trying to retrieve an email message from a backup of Lotus Notes 3 years after the organization converted to Exchange and the Notes admins left the company.