Backups Get Better

Senior Reporter, Computerworld |

Despite enormous leaps in tape subsystem and hierarchal storage management technologies for enterprise backups, most IT managers still use direct-attached tape drives that require lengthy backup windows when administrators insert and remove multiple tapes as backups run.

And as the number of low-cost Windows and Unix servers has grown, IT managers have simply added more independent tape drives, increasing management complexity.

The move toward data consolidation on storage-area networks and network-attached storage devices has enabled administrators to add automated tape libraries (ATL), which reduce backup times of these larger, consolidated data sets by using multiple tape drives to write backups concurrently to hundreds of cartridges within a single ATL.

But while tape libraries have consolidated storage, backups still shut down application servers for an average of 14 hours, according to Robert Amatruda, an analyst at IDC in Framingham, Mass. And administrators may wait tens of seconds to restore a file from tape (once it's mounted) and several minutes to retrieve several files. "It just takes forever to get to your data," Amatruda says.

Disk in the Middle

That's led to the recent trend of positioning tape as an archival solution only and using disk drives as both backup storage and as a tape buffer for archiving. The cost of Advanced Technology Attached disk drives has dropped to the point where they're cheaper than SCSI drives by a factor of three. This makes them an attractive alternative for backup storage.

Disk-to-disk backup systems support rapid data restoration at disk speeds and can function as near-line storage devices. For archival backups, the systems act as a cache between the primary storage and tape subsystem, so that the lengthy backups needed to stream to tape don't tie up the primary storage devices.

Vendors of disk-to-disk backup products, such as Sunnyvale, Calif.-based Network Appliance Inc., advocate backing up data exclusively to the backup appliance for a period of time—days or weeks—and creating less frequent tape backups for archival purposes.

"That data is on the disk for a few days," says Robert Abraham, an analyst at Freeman Reports in Ojai, Calif. "After that time, the likelihood you'd need it to restore diminishes quickly. The user most likely would take the data off the disk and put it on tape. That frees up disk space so that [the disk backup system] only has current data."

High Price Tags

Using disk-to-disk backup systems as a buffer for tape backup can eliminate the problem of long backups that tie up primary storage. But the systems aren't cheap; they can cost $275,000 or more, depending on capacity.

For companies with mainframes, the problem of tape storage is exacerbated by the fact that tape drives have been limited to holding one data set per cartridge, often leaving the majority of tape unused.

Virtual tape servers resolve that. These systems combine a tape library with a caching disk array and special storage management software that makes the array look like a tape library to the host. They can also stack multiple data sets on a single cartridge, filling each cartridge to capacity and saving space. IBM's current virtual tape server, for example, can emulate up to 256 tape drives. Virtual tape servers rely on higher-performance SCSI arrays and support both Fibre Channel and mainframe interfaces such as Escon. A typical installation may cost $250,000 or more.

"I think things like near-line and virtual storage are for a group of users that have a lot of infrastructure and a lot of money that's outlaid to help in that regard," Amatruda says. "Virtual tape is still a high-cost support-and-deploy solution. Your average systems guy doesn't have that."

If subsystem prices continue to fall, that could eventually change. "At the end of the day, you'll still find more of a blended solution. You will have disk and tape operating much closer to one another in terms of backup function," Amatruda says.

IBM's Virtual Tape Server Architecture

IBM's virtual tape server emulates a tape library but backs up data to a disk array before migrating it to tape. Other vendors use similar designs. Unlike disk-to-disk backup systems, virtual tape server systems use high-performance SCSI drives and include mainframe interfaces.

Cheap & Secure Data Stores

Stories in this report:

Data Center

Senior Reporter Lucas Mearian covers AI in the enterprise, Future of Work issues, healthcare IT and FinTech.

It’s time to break the ChatGPT habit