Ask the Expert

Disk vs. tape for long-term backup

I've enjoyed your columns on backup and recovery and even attended your seminar at a few TechTarget forums.

In my environment, I use four different applications (BackupExec, NetBackup, HPDP and plain ol' tar) to backup various platforms to various tape drives and technology (LTO-2 and SDLT220/SDLT320). I have an extremely low backup failure rate, and it has rarely been the result of failed drives -- usually bad media or the contention for an available drive.

My question is, after reading many of the new reports and articles on SearchStorage, why do you think many users are abandoning tape for long-term archival backups? Do they really expect to backup to ATA or SATA disks, put them on a shelf and then dust them off in seven or eight years and expect them to spin up? Sure, I can understand D2D for daily backups, but for long-term archives? Do they realize how much labor, power and data center space is consumed if using disk for this purpose?

    Requires Free Membership to View

    Login

Most customers are not yet abandoning tape for long-term archival backups. Most are only doing so for short-term, cyclical backups that are stored for month, not years. However, that isn't to say that's it's not a defensible idea. First, they're not (yet) talking about powering them off and putting them on a shelf. However, there are vendors (such as Copan) who are using massive array of independent disk (MAID), where disks are powered off more than they're powered on. In a Copan array, only 20% of the drives are powered on at any one time. The other 80% sleep to save power and cooling.

Every so often, each drive is powered on, parity is used to verify the data written to each drive, then it is powered back off. Drive use is tracked and data is automatically migrated from busy drives to less-busy drives, so that all drives in the array are used roughly the same amount of time. This, of course, significantly reduces the amount of power consumed and significantly increases the MTBF.

Even without MAID, I've made the argument that the cost associated with power, floor space and labor is still significantly less than the amount of power, floor space and labor required to maintain a tape-based infrastructure. I did some calculations using the biggest, fully populated disk system I could find, powered it on all the time and calculated how much power that would cost. It came out to something well under $10,000 a year. That translates into 1/5-1/10 of the cost of an FTE, and I'd argue that maintaining that disk system is a heck of a lot less expensive in labor dollars than maintaining the tape system. Regarding the footprint, disk systems beat tape systems any day on a GB-per-square-foot comparison.

This was first published in April 2006

Join the conversationComment

Share
Comments

    Results

    Contribute to the conversation

    All fields are required. Comments will appear at the bottom of the article.
    Back to top